The Characters of Palaeozoic Jawed Vertebrates

Home , Acanthodes, Antiarchi, Brochoadmones, Cephalaspis, Dinichthys, Errivaspis

bs_bs_banner

Zoological Journal of the Linnean Society, 2014, 170, 779–821. With 13 ﬁgures

The characters of Palaeozoic jawed vertebrates

MARTIN D. BRAZEAU1* and MATT FRIEDMAN2

1Naturalis Biodiversity Center, P.O. Box 9514, 2300 RA Leiden, The Netherlands 2Department of Earth Sciences, University of Oxford, South Parks Road, Oxford OX1 3AN, UK

Received 14 June 2013; revised 27 October 2013; accepted for publication 30 October 2013

Newly discovered fossils from the Silurian and Devonian periods are beginning to challenge embedded perceptions about the origin and early diversification of jawed vertebrates (gnathostomes). Nevertheless, an explicit cladistic framework for the relationships of these fossils relative to the principal crown lineages of the jawed vertebrates (osteichthyans: bony fishes and tetrapods; chondrichthyans: sharks, batoids, and chimaeras) remains elusive. We critically review the systematics and character distributions of early gnathostomes and provide a clearly stated hierarchy of synapomorphies covering the jaw-bearing stem gnathostomes and osteichthyan and chondrichthyan stem groups. We show that character lists, designed to support the monophyly of putative groups, tend to overstate their strength and lack cladistic corroboration. By contrast, synapomorphic hierarchies are more open to refutation and must explicitly confront conflicting evidence. Our proposed synapomorphy scheme is used to evaluate the status of the problematic fossil groups Acanthodii and Placodermi, and suggest profitable avenues for future research. We interpret placoderms as a paraphyletic array of stem-group gnathostomes, and suggest what we regard as two equally plausible placements of acanthodians: exclusively on the chondrichthyan stem, or distributed on both the chondrichthyan and osteichthyan stems. © 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821. doi: 10.1111/zoj.12111

ADDITIONAL KEYWORDS: Acanthodii – Chondrichthyes – Gnathostomata – morphology – Osteichthyes – palaeontology – Placodermi – plesiomorphy – synapomorphy.

[T]he evidence from all useful characters must be weighed of living taxa. Instead, the problems surround the impartially, and unbiased comparisons must be made with all placement of fossil taxa with respect to the modern potentially related groups. groups. Gnathostomes are divided into two extant R. Denison (1979: 19). lineages: the Chondrichthyes (sharks, rays and skates, and chimaeras) and much more diverse INTRODUCTION Osteichthyes (bony fishes and tetrapods). That these are sister groups and reciprocally monophyletic is The phylogenetic relationships of early fossil not greatly disputed, and is well supported by both gnathostomes (jawed vertebrates) remains one of the molecular and morphological studies (Nelson, 1969; most significant but under-researched problems in Wiley, 1979; Maisey, 1986; Takezaki et al., 2003; Blair vertebrate palaeontology. Unlike many other areas of & Hedges, 2005; Chen et al., 2012). Debate currently vertebrate systematics, the debates do not concern concerns the phylogenetic placement and monophyly how fossils might illuminate the inter-relationships of two exclusively fossil assemblages: the placoderms, armoured fishes with simple jaws and dental structures that are hotly debated homologues of teeth, and *Corresponding author. Current address: Department of Life the acanthodians, shark-like fishes with numerous Sciences, Imperial College London, Silwood Park Campus, Buckhurst Road, Ascot SL5 7PY, UK. bony spines preceding their fins. When gnathostome E-mail: [email protected] fossils cannot be placed in either of the extant groups

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd 779 on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. 780 M. D. BRAZEAU AND M. FRIEDMAN they have typically been assigned to either of these These include the relative order of appearance of extinct categories. From a phylogenetic perspective, features such as an epicercal tail, paired pectoral that leaves few fossil branches that can be used to appendages, and perichondral bone (Forey & Janvier, infer sequences of character acquisition in important 1993; Janvier, 1996a; Donoghue et al., 2000; Janvier, parts of the gnathostome tree, such as those that 2001; Gai et al., 2011). document the origin of jaws and the unique anatomi- By comparison, the application of cladistic methods cal attributes of chondrichthyans and osteichthyans. to the early gnathostome problem is in a state of Although fossils of Devonian and Silurian fishes infancy. This relative lack of progress is paradoxical: that challenge old perceptions are frequently coming early gnathostome fossils are similarly diverse as their to light (Zhu, Yu & Janvier, 1999; Maisey, 2001; jawless counterparts and almost certainly richer in Maisey & Anderson, 2001; Miller, Cloutier & Turner, characters. Only a few numerical cladistic analyses 2003; Hanke & Wilson, 2004, 2006, 2010; Brazeau, have had the taxonomic scope to test such fundamental 2009; Maisey, Miller & Turner, 2009; Zhu et al., 2009, phylogenetic questions as the monophyly of placo- 2012a, 2013), comparatively little effort has been derms and acanthodians, and evaluate competing spent to investigate their phylogenetic relationships. placements for stem members of Osteichthyes, Chon- In addition to this, the record of even the earliest drichthyes, and Gnathostomata (Friedman, 2007a; jawed vertebrates is characterized by considerable Brazeau, 2009; Davis et al., 2012; Zhu et al., 2013). In anatomical disparity (Anderson et al., 2011; Davis, this paper, we elaborate on the body of character Finarelli & Coates, 2012). Currently, placoderms information emerging from these recent analyses. We are the only group of widely accepted stem-group explore conflicting data, and frame the relevant sys- gnathostomes that are known to exhibit jaws (Young, tematic questions, as we see them, with greater explic- 1986). Acanthodians have been placed in this position itness, and set clear avenues for future research. To do at one time or another (Watson, 1937; Rosen et al., this, we address three fundamental problems: 1981; Brazeau, 2009; Davis et al., 2012), but this has 1. How might previous palaeontological collection, rarely been consistent and the character support is research, and analysis have failed to populate unclear. Meanwhile, fossils that branch between the naked stem-group branches of the chondrich- placoderms and their nearest jawless relatives have thyan, osteichthyan, and crownward parts of the yet to be discovered or identified. Consequently no gnathostome stem (Fig. 1)? anatomical intermediates are known to punctuate 2. What are the principal characters describing the this conspicuous gap. The number of stem osteich- hierarchy of these three gnathostome branches thyans and stem chondrichthyans identified with incident to the gnathostome crown node? any confidence is similarly limited, with candidates 3. What are the implications of the resulting scheme usually sourced from the acanthodians. Nevertheless, for interpreting problematic or unusual taxa? the fossils of early gnathostomes are sometimes used as background for hypotheses of character and devel- To address these questions, we need to re-evaluate opmental evolution, even though these relationships the phylogenetic status of the two main extinct groups, and character transformations remain unclear (e.g. placoderms and acanthodians, through a critical Beverdam et al., 2002; Koentges & Matsuoka, 2002; examination of their characters. As in our previous Gillis et al., 2011). work on osteichthyans (Friedman & Brazeau, 2010), Over the past three decades, pioneering cladistic we can assess whether our current systematic tradi- work has made it clear that Palaeozoic armoured tions may be causing us to misidentify or misinterpret jawless fishes, sometimes called ‘ostracoderms’, are a fossils. We feel the best approach to the problem is paraphyletic array of stem gnathostomes (Janvier, captured by Denison’s (1979) quote in the epigraph. 1981a, 1984; Forey & Janvier, 1993). This brought Denison invites us to make comparisons not just an end to decades of work that considered the within the groups that we have inherited by the ‘ostracoderms’ almost exclusively in light of them- traditions of our discipline, but to compare each selves (see review in Janvier, 1996b), and intro- one objectively and across the perceived taxonomic duced a computational framework for studying boundaries as though they did not exist at all. In this stem gnathostome relationships (Donoghue, Forey & way, we can better establish the quality of evidence Aldridge, 2000). Although uncertainty and disagree- that forms the basis of our current systematics of early ment remains about the precise relationships of gnathostomes. jawless stem gnathostomes, debates on the topic have proved highly fruitful. The resulting synapomorphic hierarchies have elucidated the step-wise acquisition A HISTORY OF THE ‘PLACODERM PROBLEM’ of a suite of anatomical features that distinguish The taxonomic history of Placodermi includes two gnathostomes from their living jawless relatives. major eras, each spanning nearly a century (Obruchev,

1964; Goujet, 1984b). The first of these corresponds to Thus within decades of its formulation, Placodermi the initial identification of the taxon, followed by serial lay in tatters. Woodward’s (1891) classification cap- rejections of its coherence and redistribution of its tures prevailing sentiments at the close of the 19th constituent members to a range of vertebrate groups. century. Coccosteus and other arthrodires, along with The second period began with the rapid reunion of Chelyophorus, were lungfishes, Pterichthyodes and most original members along with newly discovered other antiarchs were allied with ‘ostracoderms’, and groups, followed by a long period of stasis in the Psammosteus was compared with sharks. Ptyctodonts concept of Placodermi that continues to the present. (exclusive of Chelyophorus) had yet to be associated M’Coy (1848) erected Placodermi to include with placoderms, and were placed with holocephalans. members of Agassiz’ (1844) Cephalaspides minus Woodward’s scheme persisted largely unscathed in Cephalaspis with some additions: the heterostracan textbooks well into the 20th century (Woodward, 1932). Psammosteus and a series of other genera (e.g. the The modern concept of Placodermi is often attrib- ptyctodont Chelyophorus) previously aligned with coe- uted to Stensiö (1925, 1931), Gross (1931, 1937), and lacanths. These taxa, M’Coy argued, were united by Heintz (1932) (Obruchev, 1964; Goujet, 1984b), but the presence of large, tuberculated plates covering key steps in reuniting the placoderms had already the head and trunk. Pander (1857), Kade (1858), taken place in the first decade of the 20th century. and Traquair (1888) maintained a taxon closely cor- Hussakof, who had earlier (1905, 1906) reiterated responding to M’Coy’s Placodermi. Then as now, Dean’s (1899, 1901) arguments that arthrodires were arthrodires were interpreted largely in the light of unrelated to lungfishes, rejected the link between living gnathostomes. Huxley (1861: 37) argued that antiarchs and ‘ostracoderms’. Hussakof (1906: 134) Coccosteus was allied with catfishes, and in asserting asserted that these groups were ‘united on negative that ‘wherever Coccosteus goes, Pterichthys must evidence . . . rather than for the possession of a series follow . . . though the structure of the last-named fish of common characters’. Critical to Hussakof’s thesis is . . . more difficult of interpretation’, inaugurated was Patten’s (1904) discovery of mouthparts in a new tradition: the investigation of poorly known Bothriolepis comparable to but ‘inferior in develop- or unusual placoderms through an arthrodire model ment to [those] of the Arthrodira’ (Hussakof, 1906: with closer correspondences to modern gnathostomes 135). As antiarchs could no longer be dismissed as (e.g. Young, 2010: fig. 1). jawless, Hussakof argued that correspondences Huxley was not unique in aligning placoderms with between the dermal carapace of this group and osteichthyans; Newberry (1875), Crane (1877), Cope arthrodires were evidence of a close relationship. (1889, 1892), and Regan (1904) championed links Hussakof (1906) came to another prescient conclu- between arthrodires and sarcopterygians. However, sion: placoderms are not members of the extant Huxley’s enthusiasm to brush aside the differences gnathostome radiation. between arthrodires and antiarchs was not universal. By the 1930s the construction of a modern concept The latter were banished to Agnatha by Cope (1889, of Placodermi was effectively complete. The content 1892), who remarked ‘there is a wide gap between of the group varied over the coming years, including these forms and any of the fishes’ (1892: 281). Others the temporary embrace of acanthodians (Watson, simply ignored placoderms in their classifications 1937; Moy-Thomas, 1939; Romer, 1945, 1966) and (e.g. Gill, 1872) or declared the assemblage incertae invocations of placoderm paraphyly with respect to sedis (Lutken, 1871). chondrichthyans (e.g. Ørvig, 1962; Stensiö, 1963,

CYCLOSTOMATA ELASMOBRANCHIIHOLOCEPHALI SARCOPTERYGII ACTINOPTERYGII

CHONDRICHTHYES OSTEICHTHYES

GNATHOSTOMATA

Figure 1. Assumed phylogenetic framework for the principal extant clades of vertebrates used in this analysis.

1969; Jarvik, 1980), but the concept of Placodermi presence of semidentine, a specific kind of hard tissue; [5] an that stabilized prior to the Second World War remains omega-shaped palatoquadrate, with adductor muscles insert- effectively indistinguishable from that which has ing on the ventral surface of this structure and the internal persisted to the present day. It is important to recog- face of the suborbital plate; [6] fusion between dorsal elements nize that workers of the time were comfortable of the mandibular and hyoid arches with dermal plates of the with the notion of paraphyletic assemblages as legiti- cheek. mate taxonomic groups, and many authors under- Goujet (1984b: 237) was later able to expand his list stood that placoderms might be ‘united’ by shared of placoderm synapomorphies, which had grown from primitive characters rather than any specializations six to 11. These additional characters were: of their own. This sentiment was clearly stated by Moy-Thomas (1939: 29) in his influential Palaeozoic [7] endocranium composed of two ossifications (rhinocapsular Fishes: ‘the more knowledge of them [placoderms] has and postethmo-occipital) separated by a fissure, unless sec- increased the more certain it has become, that they ondarily fused; [8] long ethmoid region of the endocranium with terminal nasal capsules and a long subnasal shelf; [9] represent a large early gnathostome group, probably lateral orbits; [10] variable skull pattern, with numerous containing the ancestors of all modern fishes’. plates; [11] cheek covered by three plates, including a large The modern era of placoderm systematics began submarginal. with Denison’s (1975: 9) character-based review of placoderm intrarelationships. He concluded that Subsequent studies regarded many of these charac- the group could be recognized on the basis of: ters as uninformative for the usual reasons: they were an anteriorly placed gill chamber, lying beneath either primitive, or their polarity could not be deter- the neurocranium; a neck joint between the mined through outgroup comparison. Not long after neurocranium and synarcual; and dermal bones cov- Goujet’s argument for placoderm monophyly, Maisey ering the head and shoulder girdle. The first and last (1986: 225) concluded that support for Placodermi of these features are demonstrably plesiomorphic might be more apparent than real, and that ‘[c]harac- based on outgroup comparison, whereas the second is, terization of placoderms as a monophyletic group . . . is as Denison himself admitted, possibly homoplastic. problematical’. Miles & Young (1977) built upon Denison’s (1975) Goujet (2001: 210) later produced a list of five efforts at inferring placoderm intrarelationships, but synapomorphies [their equivalent(s) from Goujet, the placement of placoderms as a whole to other 1984b are given in parentheses]: a dermal shoulder groups of gnathostomes was their primary preoccu- girdle encircling the trunk and making an articulation pation. Like Denison, Miles & Young (1977) did with the skull through a joint (a partial combination of not focus on the question of placoderm monophyly, characters 1 and 2); a distinctive pattern of dermal and instead worked under the assumption that bones contributing to the skull roof and cheek (a placoderms formed a clade to the exclusion of other combination of characters 3 and 11); simple jaws gnathostomes. These formative works set the agenda bearing two or three pairs of bony plates (a new for subsequent investigation: the relationships within character); direct connection between the dermal oper- placoderms (e.g. Young, 1980, 1986; Gardiner, 1984a; culum and braincase via a hyoid arch cartilage (a Goujet, 1984b; Forey & Gardiner, 1986; Goujet & subset of character 6); and the presence of semidentine Young, 1995), and the relationships of placoderms (character 4). In response to cladistic analyses rooted to other vertebrates were topics for research (e.g. on jawless vertebrates that can test – but have failed to Goujet, 1982, 1984b; Gardiner, 1984a; Young, 1986), support – placoderm monophyly (Friedman, 2007a; but placoderm monophyly was not subject to critical Brazeau, 2009; a similar pattern has been found testing. subsequently by Davis et al., 2012 and Zhu et al., The most explicit arguments for placoderm 2013), Young (2010) more than trebled Goujet’s most monophyly (Goujet, 1984b, 2001; Young, 2010) share recent list. As the most recent argument favouring a a common pedigree, tracing their ancestry to Goujet’s classical Placodermi, we provide a detailed review of (1982: 29) list of synapomorphies. These are summa- this synapomorphy scheme in a later section. rized here based on our own translation: [1] dermal skeleton formed of plates contributing to a head and trunk shield, with the latter forming a complete ring and A HISTORY OF THE ‘ACANTHODIAN PROBLEM’ supporting the pectoral fins; [2] the presence of a double cervical joint with an endoskeletal component (between occipi- The concept of acanthodians as a coherent assemblage th tal condyles and a synarcual) and a dermal component (the can be traced to the mid-19 century (reviewed by posterolateral part of the skull overlapping the front margin of Heyler, 1969; Miles, 1973b; Denison, 1979). Agassiz the anterior dorsolateral plate); [3] a specific pattern of dermal described spines of fishes now called acanthodians in plates contributing to the head and trunk shields; [4] the his Recherches (1833–1844), but it was only in his

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 EARLY GNATHOSTOME CHARACTERS 783 monograph on fishes from the Old Red Sandstone extinct gnathostomes as a clade (Brazeau, 2009; Davis that he recognized a taxon with links to the modern et al., 2012; Zhu et al., 2013). concept of the group (Agassiz, 1844). Agassiz’ Acantho- diens contained Acanthodes, Cheiracanthus, and NAKED STEMS REVISITED Diplacanthus. Within this group, which clearly pres- ages modern concepts of Acanthodii, Agassiz included This paper seeks to address the long-standing failure an outlier: Cheirolepis. This actinopterygian, which to populate naked stem branches of the Osteichthyes, was later recognized as such by Traquair (1875), was the Chondrichthyes, and crownward parts of the not the only crown osteichthyan to be interpreted gnathostome stem. There are two possible explana- as an acanthodian; the sarcopterygian Onychodus was tions for this problem: the appropriate fossils are assigned to this assemblage by some workers well into either unpreserved or undiscovered; or the fossils have the 20th century (e.g. Romer, 1933, 1945). been incorrectly identified when found (Friedman & Although Agassiz regarded acanthodians as distinct Brazeau, 2010). In the second case, we are confronted from both ‘ganoids’ and sharks, the next 150 years of with two further possible causes: that the data are research on acanthodian relationships can be summa- misleading or that we have misidentified candidates rized as a series of competing hypotheses aligning through error. Although Acanthodii and Placodermi this group with either chondrichthyans (Roemer, are stalwarts of historical classification schemes, the 1857; Fritsch, 1890; Woodward, 1891; Dean, 1895, persistence of both assemblages does not necessarily 1907, 1909; Reis, 1895, 1896; Nielsen, 1932; reflect strong or even explicit evidence for their Woodward, 1932; Holmgren, 1942; Ørvig, 1957; monophyly. We contend that their longevity reflects a Nelson, 1968, 1969; Jarvik, 1977, 1980) or combination of convention and convenience, partnered osteichthyans (Kner, 1868; Zittel, 1893; Huxley in with a lack of decisive phylogenetic tests of their Davis, 1894; Nielsen, 1949; Heyler, 1958, 1962; proposed synapomorphies. This contributes not only to Romer, 1968; Schaeffer, 1968, 1969; Miles, 1973a; their persistence, but also to the positive and system- Gardiner, 1984a; Maisey, 1986), plus rare proposals atic misidentification of newly discovered fossils. drawing links with placoderms (Watson, 1937; Romer, 1945, 1966) or placing acanthodians on the gnathostome stem (Rosen et al., 1981). METHODOLOGICAL OBSTACLES TO Despite debate concerning the position of A COHERENT SYSTEMATICS OF acanthodians relative to other fishes, acanthodian EARLY GNATHOSTOMES monophyly has gone largely uninterrogated. One of In this paper, we attempt to avoid a set of key the few explicit arguments was outlined by Maisey problems that we perceive as common to many pre- (1986: 225), who presented two ‘admittedly weak’ vious attempts to resolve early gnathostome relation- characters supporting the group. With their status ships: self-referential assumptions of monophyly, as a clade accepted despite a lack of compelling ana- assembly of character lists without accompanying tomical evidence, most modern cladistic investigation phylogenetic tests, and the use of compound charac- of Acanthodii has concerned inter-relationships of ters. Here we outline the effects of these approaches its supposed members (Denison, 1979; Long, 1986; and show why we think they are misleading using Maisey, 1986; Hanke & Wilson, 2004; Burrow & concrete examples from the literature. Turner, 2010). The discovery of well-characterized chondrichthyans (Miller, Cloutier & Turner, 2003) and osteichthyans SELF-REFERENTIAL ASSUMPTIONS OF MONOPHYLY (Zhu et al., 2009) with paired fin spines, as well as The first problem arises from the way certain fossil diverse array of spiny early gnathostomes from the groups are interpreted solely in light of themselves Early Devonian Man on the Hill (MOTH) locality (e.g. Denison, 1975, 1978; Miles & Young, 1977; (Bernacsek & Dineley, 1977; Gagnier & Wilson, 1996; Goujet & Young, 1995). In instances where outgroup Gagnier, Hanke & Wilson, 1999; Hanke, Davis & comparisons are used, procedures are inconsistent or Wilson, 2001; Hanke, 2002, 2008; Hanke & Wilson, inexplicit. The phylogenetic analysis of Goujet & 2004, 2006, 2010; Hanke & Davis, 2008, 2012) have Young (1995) was admittedly an exploratory investi- reinvigorated debate on the monophyly and relation- gation, but the results are routinely re-printed for ships of the acanthodians. Some analyses have pro- use in comparative studies (Goujet, 2001; Smith & posed acanthodian monophyly and focused on the Johanson, 2003; Goujet & Young, 2004; Johanson & question of inter-relationships within this group Smith, 2005; Carr, Lelièvre & Jackson, 2010). This (Hanke & Wilson, 2004; Burrow & Turner, 2010), analysis was rooted on a hypothetical all-zero ances- whereas analyses with broader taxon samples have tor. Of the 49 characters used, at least 32 of the generally rejected the status of this assemblage of presumed primitive states were conditions that

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 784 M. D. BRAZEAU AND M. FRIEDMAN could only be observed in other placoderm taxa. That conditions that defines them, compound characters is, they had no relevant comparator in any non- imply all-or-nothing similarity in satisfying their cri- placoderms. The assumed primitive characters of teria even though there is no biological or logical rule placoderms are thus largely based on the presuppo- against homologues having partial resemblance. sition of monophyly and hypotheses about which For example, Burrow, Trinajstic & Long (2012: 349) placoderms can be viewed as primitive. proposed, based on Burrow & Turner (2010), that These problems were acknowledged by Goujet & ‘the main synapomorphy of the group [Acanthodii] is Young (1995, 2004) who presented the work as a a perichondrally ossified scapulocoracoid, with a stimulus for future research. We have decided to slender dorsal shaft widening out basally to a blade apply a different approach modelled on our earlier that articulates with the pectoral fin spine’. This investigation of osteichthyan characters (Friedman & character can be decomposed into at least five Brazeau, 2010). We presented a list of characters for separate variables: the presence of a scapulocora- hierarchically ordered groups within the osteichthyan coid (a general gnathostome character); perichondral total group, based on reference to a fixed outgroup ossification (a general gnathostome character); the arrangement. The validity of these synapomorphies presence of a dorsal shaft (a trait shared with can therefore be challenged and refuted by: (1) the chondrichthyans); a ventral widening of the basal discovery of alternative outgroupings; and (2) the part of the scapular blade (also seen in chon- discovery of taxa that establish incongruent or con- drichthyans); and its articulation with the fin spine junctive character distributions (cf. Patterson, 1982a; (such as that found in placoderms and possibly spine- de Pinna, 1991). bearing chondrichthyans and osteichthyans). Under scrutiny, this character dissolves as an acanthodian synapomorphy (Fig. 2) as its individual parts can be SYNAPOMORPHY AND ITS EXAGGERATION shown to either have wider distribution or be simply Vertebrates identified as placoderms and acanthodians irrelevant to many taxa (i.e. those where a spine is exhibit distinctive features that might reasonably absent). be offered as candidate synapomorphies. However, it is unclear that these resemblances reflect synapomorphies rather than symplesiomorphies or homoplasy. PLACODERM CHARACTERS OF YOUNG (2010): The task of systematists is resolution of these issues, A REVIEW but the construction of longer lists of ‘confirming In response to recent proposals of placoderm instances’ is only a first step towards this goal. paraphyly, Young (2010) expanded the list of proposed Phylogenetic relationships are not built on overall placoderm synapomorphies to 16. Although we similarity, but on hierarchical character distributions; agree with some of Young’s anatomical reappraisals it is these that are actual synapomorphies. The con- (e.g. the supposed passage of the subclavian artery struction of ever-longer lists of resemblances conflates through the postbranchial lamina in antiarchs; similarity with synapomorphy, and ignores (or is at Johanson, 2002), the list exhibits all of the issues least inexplicit about) the key hierarchical component described above, and results in an exaggeratedly long of the equation. This approach will tend to exaggerate list of synapomorphies. the number of synapomorphies, even if the group does turn out to be a well-corroborated clade. Exaggerated 1. Distinctive pattern of dermal bones in skull roof, lists of synapomorphies will ultimately lead to the cheek, and operculum. systematic misidentification of newly discovered This is too vaguely stated to be subject to any specific species. test. However, it is indeed the case that the cranial bones of placoderms are identified in terms of those COMPOUND CHARACTERS bones in other placoderm species under the assumption of a symplesiomorphic pattern. The resulting Compound characters reflect a combination of char- ‘distinctive’ pattern is therefore unsurprising, and this acter particles (conditions) assembled into one state- itself has recently been challenged directly (Zhu et al., ment to give the appearance of one character. 2013). Without an inferable plesiomorphic condition, Morphologists understandably seek to give greater the distinctiveness of placoderm skull bone patterns is precision to their characters by composing multiple moot. Only osteichthyans provide a possible outgroup conditions to identify them. Unfortunately, this fails condition, leaving the possible primitive condition to distinguish compatibility from congruence, and equivocal. leaves aside the possibility that the compounded conditions may individually have greater levels of 2. Paired external openings of endolymphatic ducts phylogenetic generality. As it is the combination of linked to dermal bone ossification centres (nuchal

A + Parexus + Acanthodes CHONDRICHTHYES Climatius Ischnacanthus OSTEICHTHYES

Perichondral bone

zero-length branch Ventral widening of scapulocoracoid Dorsal scapular blade Presence Perichondral bone Scapulocoracoid Absence Applicable upwards only (= no cost)

Parexus Acanthodes Ischnacanthus CHONDRICHTHYESClimatius OSTEICHTHYES

Perichondral bone

Ventral widening of scapulocoracoid Dorsal scapular blade Perichondral bone Scapulocoracoid

Figure 2. Decomposition of a compound character. The proposed synapomorphy of acanthodians (Burrow & Turner, 2010) can be decomposed into separate conditions each having greater levels of generality. A, acanthodians as a monophyletic group and the compound character of Burrow & Turner yields a zero-length branch. B, collapsed Acanthodii, showing true level of support for the individual character components.

or paranuchal plates) in posterior part of the do not pass through cranial plates. All non-placoderms skull roof. in which there are endolymphatic openings in the skull roof are either tessellate, micromeric (osteostracans This is a compound character that is conditioned on and acanthodians: Miles, 1973a; Sahney & Wilson, the following criteria: there are any endolymphatic 2001), or bear a single large shield (some osteo- openings in the skull roof at all; there are distinc- stracans and galeaspids; Janvier, 1985a, b, 1996a; Gai tive skull roofing bones identified as nuchals or et al., 2011), precluding any a priori attempt to resolve paranuchals. This character cannot be considered plesiomorphy from apomorphy. independent evidence from character 1 because the identification of a nuchal plate is considered part of the 3. Simple jaws with only two or three pairs of bony distinctive placoderm skull roof condition. To condition tooth plates. this character on the passage of ducts through a nuchal or paranuchal plate would presuppose that some non- There is no clear reason why simple jaws cannot (or placoderms have nuchals or paranuchals. The only should not) be a gnathostome symplesiomorphy. As non-placoderms with macromeric skull roofing bones with the pattern of skull roof bones, the variability of are osteichthyans, in which the endolymphatic ducts outgroups makes it difficult to establish whether this

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 786 M. D. BRAZEAU AND M. FRIEDMAN condition is primitive or derived. Chondrichthyans position on its dorsal face (cf. multiple coronoids of and most acanthodians have no large tooth-bearing osteichthyans) and lacking external and internal dermal bones, whereas osteichthyans have several. cover of dermal bones (cf. dentary, prearticular of However, two pairs of bony plates (upper and lower) osteichthyans). are found in the jaws of some acanthodians including Tetanopsyrus (Hanke et al., 2001) and ischnacanthids This arrangement is not unique to placoderms, (Burrow, 2004). because the same character state is apparent in Tetanopsyrus (Hanke et al., 2001) and ischnacanthids 4. Omega-shaped upper-jaw cartilage (palatoqua- (Burrow, 2004). In addition to being compound, this drate), with deep ventral embayment for character repeats much of the information given in adductor mandibulae muscle (which was not character 3, and therefore these two characters are not enclosed mesially by dermal bone; see 6, 8). independent. Because all placoderms differ from the only speciﬁed outgroup (in this case, osteichthyans), Others (e.g. Schaeffer, 1975; Janvier, 1996a) have then again, the polarity of the precise number of plates already noted that it is not possible to establish this is unknown. character as a placoderm synapomorphy as opposed to a gnathostome symplesiomorphy without making 10. Adductor mandibulae muscle with broad lateral the prior assumption that placoderms are crown- attachment to Meckel’s cartilage (cf. posteriorly group gnathostomes. conﬁned adductor fossa of osteichthyans).

5. Palatoquadrate enclosed by lateral attachment to This assertion about the comparative breadth of one or two dermal bones (suborbital, postsubor- muscle insertion area in osteichthyans (or any bital plates), neither carrying socketed teeth (cf. other gnathostomes) and placoderms is misleading, maxilla, quadratojugal of osteichthyans). conflating the area of attachment with the size of the adductor fossa (which is only known in fossil This is not a feature shared by all placoderms osteichthyans in which it is bounded by dermal (e.g. ptyctodonts, and possibly petalichthyids; Miles & bones). Although the adductor fossa in osteichthyans Young, 1977; Long, 1997), and is contingent upon is restricted to a posterior position, this does not the presence of dermal cheek plates. Amongst non- necessarily reflect the area of muscle attachment. In placoderms, only osteichthyans and Culmacanthus crown gnathostomes this insertion area is quite (Long, 1983) have ossified cheek plates. The broad. In Amia (Allis, 1897) and Chlamydoselachus palatoquadrate and its relationship to the large cheek (Allis, 1923) the area of attachment spans nearly the plate are unknown for the latter. In Mimipiscis, the entire length of Meckel’s cartilage. The adductor palatoquadrate is fused to the preopercular plate narrows in lateral perspective at the level of the jaw (Gardiner, 1984b), meaning that this condition is also joint where upper and lower members join at the found outside of placoderms. mid-lateral raphe (Wilga, 2005).

6. Mesial surface of palatoquadrate lacking dermal 11. A special type of opercular suspension, compris- bone cover (cf. entopterygoid of osteichthyans). ing a dermal submarginal plate connected directly to the braincase via a cartilage of pre- This character only implies that placoderms are not sumed hyoid-arch derivation. crown-group osteichthyans. The hyoid system is known in few placoderms, and 7. Palatoquadrate carrying single dermal gnathal in those in which any aspects are known, it is usually element (posterior supragnathal) (cf. dermopala- incomplete (Young, 1986; Trinajstic et al., 2012). tine, ectopterygoid of osteichthyans, both of which Given the absence of any kind of opercular system in carry true teeth) jawless ﬁshes, this character can only be polarized on This character is a subcondition of Young’s charac- the assumption that placoderms are crown-group ter 3, repeating some of the same information. Addi- gnathostomes. Otherwise, it might simply be a tionally, Tetanopsyrus and ischnacanthids show this gnathostome symplesiomorphy. putative placoderm trait. 12. Extensive postorbital endocranial processes fused 8. No dermally enclosed adductor fossa in upper jaw. to the inner side of the dermal skull roof to delineate muscle attachments for operculum, vis- See arguments on character 6. ceral arches, and shoulder girdle.

9. Lower-jaw cartilage (Meckel’s cartilage) carrying This is an extremely elaborate compound character. one dermal bone (infragnathal) with a primitive It relies on conditions inapplicable to many possible

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 EARLY GNATHOSTOME CHARACTERS 787 outgroups. The posterolateral vacuities identified by character therefore has no clear polarity given any Young (e.g. 1978, 1980) as the cucullaris fossae may outgrouping for placoderms. be alternatively interpreted as ‘parabranchial fossae’ 16. Special hard tissue (semidentine) in surface layer (Elliott & Carr, 2010). Regardless, a cucullaris muscle of dermal elements. is undefined for outgroups without a separation between the head and shoulders (e.g. osteostracans). Like Davis et al. (2012), we retain this character Other ridges and processes are associated attach- below but note that it remains problematic. ments of visceral arches, bringing the comparisons Semidentine is a tissue typology that can be further into line with the braincases of osteostracans atomized into distinct traits: polarization of cells, (Janvier, 1985a, b). and whether those cells are embedded in matrix. In effect, it is morphologically intermediate between 13. Branchial chamber confined beneath braincase meso- and orthodentine. There are also questions by anteroventrally sloping dermal postbranchial about the distribution of this tissue, as scales from lamina Siberia attributed to the earliest known acanthodians The branchial chamber lies beneath the braincase (Karatajute-Talimaa & Smith, 2003) and a ptera- in osteostracans (Janvier, 1985a, b), galeaspids spidomorph from the Ordovician of Australia (Halstead, 1979; Gai et al., 2011), pituriaspids (Young, (I. Sansom et al., 2013) bear semidentine-like tissue 1991), heterostracans (Janvier & Blieck, 1979), and with polarized cell spaces. at least partially in osteichthyans (Jarvik, 1980; Giles, Rücklin & Donoghue (2013) noted that not all Gardiner, 1984b). We see no alternative but to inter- placoderms exhibit a superficial dentinous layer. pret this character as a primitive gnathostome Semidentine therefore becomes a logically impossible trait (see also Zangerl, 1981). The remainder of this trait variable for a large number of placoderm taxa. character description refers to compound features Nevertheless, the broad absence of dentinous tissue predicated on a series of conditions that can be does not necessarily count against the status of atomized and treated independently. For instance, semidentine as a placoderm synapomorphy. We have postbranchial laminae are found in osteichthyans, therefore provisionally kept this character as a poten- and also slope anteroventrally. tial placoderm synapomorphy. In spite of this impressively long list of compatible 14. Exoskeletal shoulder girdle including one or two similarities in placoderm taxa (Young, 2008, 2010), we median dorsal elements overlapped with inter- cannot agree that they provide any a priori evidence locking lateral and ventral plates to form a rigid of placoderm monophyly either collectively or individ- ring encircling the trunk. ually. The status of all of these characters depends on Median dorsal plates are known in at least one their distribution in a cladogram that must be cor- early osteichthyan, Guiyu (Zhu et al., 2009), whereas roborated by other characters – a necessary back- a rigid ring of dermal shoulder elements encircling ground that has not been supplied. the trunk is found in pituriaspids (Young, 1991) and many osteostracans (Janvier, 1985a), but there is some ambiguity in this latter group. Ateleaspis METHODS AND ASSUMPTIONS (Ritchie, 1967) and Superciliaspis (Adrain & Wilson, Here we justify our assumed phylogenetic backbone 1994) lack a significant ventral pectoral girdle, in and outline the methods and assumptions used in this contrast with what could be a more specialized con- work. The goal is to list a series of highly congruent dition (Janvier, 1985a; Sansom, 2009) in taxa such as and easily identified characters that require the Norselaspis (Janvier, 1981b). Nevertheless, there fewest number of phylogenetic assumptions in addi- seems to be no way to establish this as a placoderm tion to our proposed phylogenetic backbone. synapomorphy a priori.

15. Dermal articulation between skull and shoulder TAXONOMIC AND NOMENCLATURAL CONVENTIONS girdle localized to paired dermal neck-joint In order to make our hypotheses about the inter- between anterior dorsolateral and paranuchal relationships of early jawed vertebrates completely plates. explicit, and therefore open them to direct testing and This character presupposes the macromeric con- refutation, we adopt a standardized taxonomic termi- dition as an outgroup state, and thus the only non- nology throughout this paper. Specifically, we apply placoderms that might root this character are crown-, total-, and stem-group conventions to avoid osteichthyans. However, if character 1 is accepted the systematic ambiguities that are sometimes as a placoderm synapomorphy, then osteichthyans associated with discussions of the affinities of fossil are precluded from comparison by definition. The jawed vertebrates (e.g. the identification of some

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 788 M. D. BRAZEAU AND M. FRIEDMAN acanthodians as ‘putative chondrichthyans’; Sahney CHARACTER ARGUMENTATION AND HOMOLOGY & Wilson, 2001; Hanke & Wilson, 2010; Hanke, Wilson & Saurette, 2013). Fig. 1 shows the nomen- Although we base many of our conclusions on the clatural and phylogenetic scheme. The osteichthyan outcomes of recent cladistic analyses (Brazeau, crown group comprises the last common ancestor of 2009; Davis et al., 2012; Zhu et al., 2013), this paper living bony fishes, plus all of its descendants. The derives a verbal character list without an associated osteichthyan total group comprises the crown plus all tree search. Our choice of characters to list here fossil taxa more closely related to it than any other is guided by the same criteria as those used by living group. The chondrichthyan crown group com- Friedman & Brazeau (2010). We are using a strictly prises the last common ancestor of Holocephali and outgroup-based approach and have limited our Elasmobranchii. For the sake of precision, we choose characters to those with clear binary symmetry or to apply a restrictive use of the term Elasmobranchii, for which we can resolve ambiguity with reference taking the elasmobranch crown group to include to the assumed backbone phylogeny (discussed the last common ancestor of living sharks and rays, below). We have argued synapomorphies in terms plus all of its descendants both fossil and living (see of the transformational hypotheses on which they Maisey, 2012 for a historical review of the term elas- depend. This is left opaque, or at best implicit, in mobranch). The elasmobranch total group is therefore most character lists that are not associated with a the crown plus all taxa more closely related to it than computer-based tree search. We have therefore paired to any other extant group. This stands in contrast to all arguments of synapomorphy with an explicit the way in which elasmobranch is sometimes applied transformational hypothesis from a particular plesio- as a term for any chondrichthyan with a shark-like morphic starting condition to another specific derived body-plan. The chondrichthyan total group includes condition. the chondrichthyan crown plus all fossil species more We apply two complementary approaches to homol- closely related to it than any other extant group. The ogy argumentation in this paper. The first is that all gnathostome crown includes the last common ances- hypotheses of homology are conditional (Bock, 1969). tor of Chondrichthyes and Osteichthyes plus all of its The second is that congruence constitutes the only descendants. The gnathostome total group includes real test of homology (see Patterson, 1982a). By ‘real the gnathostome crown plus all extinct taxa more test’ we mean that the test of congruence provides closely related to it than any other living group. An explicit, objective criteria for deciding when a char- important implication of this terminological scheme is acter shared between two or more species should be that we regard many jawless vertebrates as members considered nonhomologous. We leave aside hopeless of the gnathostome total group because they are more arguments that appeal to the overall similarity of closely related to living jawed vertebrates than they traits to establish the supposed ‘strength’ of a homol- are to any other extant radiation (Forey & Janvier, ogy hypothesis. They have neither explicit nor 1993; Donoghue et al., 2000). implicit criteria for when a hypothesis of homology It is into this systematic framework that we intro- should be considered disconfirmed. They arise from duce the two assemblages of extinct gnathostomes failure to explicitly express the conditions of a homol- that are the focus of this contribution: acanthodians ogy proposition and rest on arbitrary essentialistic and placoderms. We apply these terms in their tradi- definitions of terms (cf. debates on the homology of tional capacity (e.g. Moy-Thomas & Miles, 1971; teeth). As noted by Bock (1969), degrees of similarity Denison, 1978, 1979; Janvier, 1996a). Our use of these can reflect degrees of relatedness. For that reason, we terms should not be taken as an endorsement of the employ the conditional specifier ‘homologous as ...’ monophyly of the assemblages of species that they whenever there is ambiguity about the taxonomic and describe; they instead reflect hypotheses of monophyly comparative level of a hypothesis of homology. In that that can be tested and potentially rejected. When way, homologues can be proposed and refuted on these fossil groups are considered within the termino- different levels (e.g. wings of birds and bats: they are logical framework proposed above, relevant questions homologous as tetrapod limbs; nonhomologous as concerning these assemblages are rendered clear. kinds of wings). Are the placoderms or the acanthodians, as they An exception to our reliance on congruence is when are generally conceived, a clade? If yes, what are characters fail the logical test of conjunction (see the specific attributes (i.e. synapomorphies) that Patterson, 1982a). Young (2008, 2010) provided a clear unite their constituent taxa to the exclusion of example of such a refutation for the passage all other species? Regardless of their status as a of the subclavian artery as a character excluding clade, to which stem do these assemblages belong: antiarchs from placoderms and all other mandibulate chondrichthyan, osteichthyan, gnathostome, or any gnathostomes. This character was first proposed by and all of the three? Johanson (2002), and then uncritically recycled in data

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 EARLY GNATHOSTOME CHARACTERS 789 matrices by Friedman (2007a) and later Brazeau osteichthyans, placoderms, and acanthodians. This (2009) in modified form. Young (2008) showed that pattern is explained as a loss in chondrichthyans, the brachial vascularization of antiarchs does not in with cellular bone representing a synapomorphy of fact pierce the postbranchial lamina, but instead a mandibulate gnathostomes and osteostracans. secondary ridge termed the crista internalis. The Outside of osteostracans, the most cited alterna- crista internalis cannot be interpreted ad hoc as a tives for the closest jawless relatives of gnathostomes postbranchial lamina to save this homology, because are anaspids and thelodonts. Thelodonts have been the two occur in conjunction. The character can thus be proposed as a gnathostome sister group (Turner, eliminated from further consideration in this work and 1991), or even a grade with respect to jawed verte- as a potential refutation for placoderm monophyly. brates (Wilson & Caldwell, 1998). This hypothesis emphasizes their monocuspid, ‘placoid-like’ scales with a basal pore: morphology evocative of that found THE JAWLESS SISTER GROUP OF in chondrichthyans. Thelodonts have paired append- MANDIBULATE GNATHOSTOMES ages that have a controversial interpretation. The Identification of the immediate outgroup to jawed most conservative approach for our present investi- vertebrates is the first step to addressing the place- gation is to accept these structures as primary ment and testing the monophyly of problematic assem- homologues (sensu De Pinna, 1991) of gnathostome blages of early gnathostomes. We regard osteostracans and osteostracan paired appendages (i.e. they should as the jawless sister group of mandibulate gnatho- be coded the same way in cladistic matrices; see also stomes. This is supported by the presence of: (1) paired Wilson, Hanke & Märss, 2007). Thelodonts, like appendages (Fig. 3); (2) an epicercal tail (Fig. 4); (3) gnathostomes, also bear pharyngeal denticles (Van perichondral mineralization; and (4) cellular bone der Brugghen & Janvier, 1993; Smith & Coates, 2000, (Donoghue, Sansom & Downs, 2006). Characters 1, 2, 2001; Rücklin et al., 2011). However, the small size of and 3 are found in all mandibulate gnathostome these structures, the state of preservation of many lineages and therefore unequivocally plesiomorphic early vertebrate fossils, and the quality of preparation for the crown-group node. Cellular bone is absent in needed to reveal them suggests that determining chondrichthyans (see discussion below) but found in absence of this trait is difficult (Brazeau, 2012).

Figure 3. Pectoral fins of stem and crown gnathostomes. A, Errivaspis waynensis, NHMUK P.17477, a heterostracan lacking paired fins. B, Hemicyclaspis murchisoni, NHMUK P.8816, an osteostracan with paired fins. C, Cladoselache sp., NHMUK P.9276, a crown gnathostome and chondrichthyan with paired fins. Scale bars = 10 mm.

Figure 4. Tail geometry of stem and crown gnathostomes. A, Errivaspis waynensis, NHMUK P.17477, a heterostracan. B, Birkenia sp., NHMUK P.42020 (image reversed), an anaspid. C, ‘Cephalaspis’ powriei, NHMUK P.670, an osteostracan. D, Promesacanthus eppleri, UALVP 42652, an acanthodian. Scale bars = 10 mm.

Most phylogenetic analyses do not recover sister- plates; e.g. Superciliaspis, Dineley & Loeffler, 1976; group relationships between thelodonts and mandi- Adrain & Wilson, 1994). bulate gnathostomes (Forey & Janvier, 1993; Two other groups of jawless fishes must also Donoghue et al., 2000; Donoghue & Smith, 2001; Shu be considered here: Galeaspida and Pituriaspida. et al., 2003; Gess, Coates & Rubidge, 2006), as they Galeaspids resemble osteostracans in overall appear- share few features that cannot be shown to be gener- ance, but have paired nasal capsules (amphirhini) and alized traits of the gnathostome total group. We are a buccohypophyseal opening in the mouth (Gai et al., concerned that thelodont/gnathostome sister-group 2011). In these respects, they resemble gnathostomes hypotheses arise from characters given a privileged to the exclusion of other jawless fishes (but see evi- status because they are found in some shark-like dence for paired nasal capsules in pteraspidomorphs; chondrichthyans, perhaps under the supposition that a Janvier & Blieck, 1979; Gagnier, 1993), including shark-like form is primitive for jawed vertebrates as a osteostracans. They are an important complement to whole (Friedman & Brazeau, 2013). osteostracans and we consider their morphology where Maisey (1986) alternatively outlined characters possible. Nevertheless, galeaspids lack cellular bone uniting anaspids and mandibulate gnathostomes to (although the endocranium is mineralized) and paired the exclusion of osteostracans. Most of these charac- pectoral fins. Pituriaspids (Young, 1991) also resemble ters are either indirect proxies for other characters osteostracans, particularly in having paired post- (horizontal septum, fin radials), are not generally branchial fenestrae, most reasonably interpreted as found in gnathostomes (gular plates, dermal fin rays; articulation sites for pectoral fins. Paired posterior Friedman & Brazeau, 2010), or do not exclude other ventrolateral projections of the carapace of pituri- jawless fishes such as osteostracans (circumorbital aspids even hint at the possibility of pelvic fins.

Unfortunately, pituriaspids are known only as natural osteostracans are sister groups. All other jawless moulds and so no precise details of their hard tissues vertebrates are taken to be more distant relatives can be considered. Aspects of their endocranial of crown gnathostomes. anatomy arepreserved in the moulds, but there is some uncertainty of interpretation. INSTITUTIONAL ABBREVIATIONS We are not dismissing the importance of any agnathan group for understanding the characters NMS, National Museums of Scotland; NHMUK, of early gnathostomes. Furthermore, we emphasize Natural History Museum, UK; UALVP, University of that none of the ‘ostracoderms’ – including our Alberta Laboratory of Vertebrate Paleontology. selected outgroup – can serve as a surrogate ancestor for jawed vertebrates. Each group of armoured THE CHARACTERS OF GNATHOSTOMES agnathans contributes to our understanding of this problem in complementary and significant CHARACTERS THAT CAN PLACE A TOTAL GROUP ways. Nevertheless, preference for alternative, non- MEMBER CROWNWARD OF SOME ‘OSTRACODERMS’ osteostracan sister groups seems to stem from dis- The following characters place a vertebrate within the satisfaction with osteostracans as proxy gnathostome gnathostome total group, to the exclusion of a series of ancestors. For instance, Goujet & Young (1995) con- jawless groups: anaspids, heterostracans, and thelo- sidered but dismissed osteostracans as a useful donts. In practice, these features unite galeaspids, placoderm outgroup, arguing that these taxa are so osteostracans, and possibly pituriaspids, or some divergent as to preclude sound comparative study. subset of these, with jawed vertebrates. However, subsequent authors (Janvier, 1996a, b; Brazeau, 2009) have shown that this is not the Primary skeleton case. As Janvier (1996b: 269) noted, it is remarkable Endoskeletal mineralization is widely distributed that Stensiö (1925, 1969) was unimpressed by the amongst vertebrates, including jawed forms similarities between the placoderm Macropetalich- (Donoghue et al., 2006) and several jawless groups: thys and osteostracans, possibly dismissing shared Eriptychius (Denison, 1967; Smith & Hall, 1990), characteristics as general craniate features. We Euphanerops (Janvier & Arsenault, 2002), osteo- elaborate on the importance of these shared similar- stracans (Janvier, 1985a, b), galeaspids (Halstead, ities below. 1979; Zhu & Janvier, 1998, Wang et al., 2005), and quite possibly pituriaspids (Young, 1991). Extensive mineralization of the braincase is, however, limited PHYLOGENETIC BACKBONE ASSUMPTIONS to the latter three agnathan groups plus jawed vertebrates. We can combine the above information with progress Perichondral ossification (Fig. 5) of the endoskel- on the cladistic relationships of modern gnathostomes. eton is a feature shared by osteostracans and jawed We see broad agreement between molecular and mor- vertebrates (Donoghue et al., 2006). The extensive phological studies of early gnathostome interrela- endoskeletal mineralization of galeaspids appears tionships (Nelson, 1969; Wiley, 1979; Maisey, 1986; to be composed of calcified cartilage, a more general- Takezaki et al., 2003; Blair & Hedges, 2005; Chen ized tissue within vertebrates (Wang et al., 2005). et al., 2012) as an ideal starting point for debating the Pituriaspids are only known as mouldic fossils relationships of problematic fossils. We therefore posit (Young, 1991), so the nature of the hard tissues sur- the following assumptions in order to derive our con- rounding the neurocranium in this group remains clusions (Fig. 1). unknown. However, impressions of structures that 1. Gnathostomata is a monophyletic group and are at least partially interpretable in terms of an its crown group comprises Osteichthyes and endocast suggest that there was probably some Chondrichthyes. endocranial mineralization in pituriaspids as well. 2. Chondrichthyes (comprising total groups Elasmo- Osteostracans and jawed vertebrates share, to branchii and Holocephali) and Osteichthyes the exclusion of galeaspids and all other agnathans, (comprising total groups Sarcopterygii and perichondral mineralization of the sclerotic capsule. Actinopterygii) are reciprocally monophyletic, and Within jawed vertebrates, the ossified capsule is most thus constitute each others’ extant sister groups. clearly present in some placoderms (e.g. Burrow, Neither is ‘more basal’ than the other. Jones & Young, 2005; Young, 2008). The ‘sclerotic 3. Osteostracans are the sister group of all rings’ of actinopterygian osteichthyans also appear mandibulate gnathostomes. We consider conclu- to represent endoskeletal mineralization of the sions drawn in this paper to be robust to an cartilaginous sclerotic capsule (Gardiner, 1984b; alternative scenario in which galeaspids and Franz-Odendaal, 2011). Outside of osteostracans and

Figure 5. Endoskeletal mineralization of gnathostomes. A, Buchanosteus confertituberculatus, NHMUK P.48675, an arthrodire placoderm. Fractured postorbital process/lateral commissure showing perichondral lining of canals, but absence of endochondral ossification. B, Griphognathus whitei, NHMUK P.52574, a crown osteichthyan and crown sarcopterygian. Ethmoid region showing perichondrally lined canals for olfactory tracts, surrounded by endochondral ossification. C, Tristychius arcuatus, NHMUK P.57305/6, a crown chondrichthyan and stem elasmobranch. Fragment of cranial skeleton showing prismatic calcified cartilage. D, Helodus simplex, NHMUK P.8212, a crown chondrichthyan and stem holocephalan. Basicranial region showing prismatic calcified cartilage. Scale bars=5mm. jawed vertebrates, mineralized sclerotic capsules interpreted as a character uniting these groups are present in the Ordovician arandapsid Sacabam- (Janvier, 1984, 2001). Such ossifications are absent in baspis (Gagnier, 1993), a taxon generally allied anaspids (Blom & Märss, 2010), thelodonts (Märss, with heterostracans (Sansom, Donoghue & Albanesi, Turner & Karatajute-Talimaa, 2007), galeaspids (Gai 2005). If placements of arandaspids as amongst et al., 2011), and heterostracans (Janvier, 1996a). the most distant stem gnathostomes from the Outside of jawed vertebrates and osteostracans, scle- gnathostome crown are correct, the mineralized sclerotic rings have only been reported in Sacabambaspis rotic capsules of Sacabambaspis and jawed verte- (Gagnier, 1993). As with the presence of perichondral brates plus osteostracans are most parsimoniously mineralization of the sclerotic capsule, the presence of interpreted as convergent. a dermal sclerotic ring in Sacabambaspis is possibly convergent with examples in jawed vertebrates and Dermal skeleton osteostracans. Dermal sclerotic rings are found in osteostracans (e.g. Cellular bone is characterized by spaces for Ritchie, 1967; Janvier, 1985a, b) and representatives osteoblasts within the bony matrix. It is present of all traditional divisions of jawed vertebrates in osteichthyans, acanthodians, placoderms, and (placoderms: Denison, 1978; acanthodians: Denison, osteostracans (Donoghue et al., 2006; Sire, Donoghue 1979; Burrow et al., 2011; chondrichthyans: Maisey, & Vickaryous, 2009), but absent in chondrichthyans 2007; osteichthyans: Jarvik; 1980), and have been (see Sire et al., 2009; Giles et al., 2013). Cellular

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 EARLY GNATHOSTOME CHARACTERS 793 bone is usually absent in anaspids, thelodonts, and 1980; Zangerl, 1981), as well as many osteostracans pteraspidomorphs. However, I. Sansom et al. (2013) (Heintz, 1967; Ritchie, 1967; Adrain & Wilson, 1994; recently described cellular dentine in isolated micro- Keating, Sansom & Purnell, 2012) and thelodonts vertebrate remains attributed to pteraspidomorphs. (Caldwell & Wilson, 1995). However, the polarity of Perichondral bone identified in the Carboniferous this character is complicated by the presence of dorsal chondrichthyan Akmonistion is described as acellu- fins in lampreys (Marinelli & Strenger, 1954). lar (Coates et al., 1998). Bone has been identified in An additional fin character unites some agnathans Scyliorhinus (Peignoux-Devill, Lallier & Vidal, 1982) with jaw-bearing forms to the exclusion of other and described as acellular. However, these same vertebrates: paired pectoral appendages (Fig. 3). studies mention the presence of osteocytes within the Amongst jawless fishes, pectoral fins and their inter- bone matrix. Nevertheless, some doubts have been nal skeletons are best known in osteostracans raised about the identification of these tissues in (Janvier, Arsenault & Desbiens, 2004; the absence of extant chondrichthyans as bone (Clement, 1992). If such structures in tremataspids is generally inter- the identification of bone in chondrichthyans is preted as secondary; Janvier, 1996a; Sansom, 2009). correct, the acellularity of bone in chondrichthyans Large fenestrations in the dermal carapace, located would be an anatomically generalized trait, not in a position comparable to the pectoral fins of limited to the exoskeleton. osteostracans and bounded anteriorly by spine-like projections, represent circumstantial evidence for Fins similar appendages in pituriaspids (Young, 1991). Hypocercal tails (Fig. 4) are found in most jawless Remains of the fins themselves, however, are fishes: lampreys (Marinelli & Strenger, 1954), unknown. euconodonts (Donoghue et al., 2000), anaspids (Blom & Paired fins are also present in Euphanerops, Märss, 2010), heterostracans (Pradel et al., 2007; anaspids, and some thelodonts. Euphanerops is Mark-Kurik & Botella, 2009), and thelodonts (Märss unique in having what appears to be a paired anal fin, et al., 2007) (for a review, see Pradel et al., 2007). or at least an anal fin supported by paired cartilages Caudal-fin structure remains unknown in pituriaspids (R. Sansom et al., 2013). Anaspid paired fins differ and in galeaspids; it is unclear whether the hetero- considerably from the short-based appendages cercal tails of galeaspids (e.g. Sanqiaspis, Liu, 1975) common to pituriaspids, osteostracans, and jawed are are hypo- or epicercal. Osteostracans are the only vertebrates. The long-based anaspid examples extend agnathans in which an epicercal tail has been reliably the length of the flank from their anterior insertion identified, and it is therefore a feature that unites at the rear of a triradiate postbranchial spine to them with mandibulate gnathostomes (Fig. 4). their posterior termination near the level of the Complementing the epicercal caudal fin of jawed anal fin (Blom & Märss, 2010). By contrast, the vertebrates is the anal fin (Fig. 4). In osteostracans, paired appendages of the thelodont Turinia (see a narrow lobe that is closely applied to the caudal Donoghue & Smith, 2001) do not differ appreciably in lobe is comparable to the anal fin of jawed verte- either gross form or position from those found in brates. In jawed vertebrates, these fins are separated early osteostracans (e.g. Ateleaspis; Ritchie, 1967). by a broad gap or a deep notch. However, notable Character-mapping exercises using most hypotheses exceptions include certain ‘acanthodians’, such as of agnathan inter-relationships (Donoghue et al., Brochoadmones (Hanke & Wilson, 2006). Outside 2000; Gess et al., 2006) indicate that the paired of these two groups, anal fins are only known in appendages of anaspids are convergent with those of anaspids and Euphanerops, in which they are osteostracans and jawed vertebrates, but are equivo- combined with a hypocercal caudal fin (Blom & cal concerning the status of thelodont pectoral fins Märss, 2010). There is no evidence of anal fins in relative to those found in gnathostomes. heterostracans (Janvier, 1996a) or arandaspids (Pradel et al., 2007). Extant agnathans are frequently Summary considered as lacking anal fins (Marinelli & Strenger, We regard the following as features that unite some 1954, 1956), but anal fins in specimens of lamprey groups of armoured jawless vertebrates with jawed have been reported (Vladykov, 1973; Vladykov & Kott, vertebrates to the exclusion of other ‘ostracoderms’. 1980). The condition in pituriaspids and galeaspids cannot be assessed owing to a lack of suitable 1. Perichondral bone (Fig. 5A, B). postcranial material, whereas anal fins are variably 2. Endoskeletal mineralization of sclerotic capsule. present in thelodonts (Märss et al., 2007). Dorsal fins 3. Epicercal caudal fin (Fig. 4C, D). are characterized by a similar pattern of distribution, 4. Anal fin (Fig. 4B–D). being found in early representatives of all groups 5. Paired fins (Fig. 3B, C). of jawed vertebrates (Denison, 1978, 1979; Jarvik, 6. Dermal sclerotic ring.

CHARACTERS THAT CAN PLACE A FOSSIL IN THE process’ of certain arthrodires (e.g. Kujdanowiaspis, GNATHOSTOME TOTAL GROUP, CROWNWARD see Stensiö, 1963; Goujet, 1984a) as potentially OF OSTEOSTRACANS equivalent. A postorbital process is apparently absent in most sarcopterygians (Jarvik, 1980). The equiva- Neurocranium lent position in most ﬁsh-like sarcopterygians exclud- Jawed vertebrates share a series of neurocranial ing dipnoans (Miles, 1977; Friedman, 2007b) is features unknown in proximate agnathan outgroups. marked by the intracranial joint. A structure termed In shark-like chondrichthyans (Maisey, 2005, 2007; the postorbital pillar is described in some stem Maisey et al., 2009), actinopterygians (Gardiner, sarcopterygians and Styloichthys (Yu, 1998; Zhu, Yu 1984b), and Acanthodes (Miles, 1973b; Davis et al., & Ahlberg, 2001; Zhu & Yu, 2002). Although not as 2012), the neurocranium exhibits a well-developed pronounced as in other gnathostomes, the postorbital lateral projection where the orbits meet the otic pillar is a projection anterior to the spiracular region capsules (Figs 6, 7), termed the postorbital process. that forms a posterior boundary to the orbit, straddles The process termed an ‘anterior postorbital pro- the jugular vein, and joins the basipterygoid articu- cess’ in arthrodires (Goujet, 1984a) and some other lation ventrally. It can be related to the postorbital placoderms (Stensiö, 1969; Ørvig, 1975) supports process on the basis of multiple conditions. We there- the hyoid arch articulation. Its anterior surface therefore consider the postorbital process a generalized fore delimits the posterior boundary of the spiracular feature of crown-group gnathostomes. This argument chamber and it cannot be interpreted as a postorbital echoes the identiﬁcation of a postorbital pillar in process equivalent to those of osteichthyans and Entelognathus by Zhu et al. (2013). chondrichthyans, in which the extension is anterior to Orbital morphology of most crown gnathostomes the spiracular space. Rather, we see the ‘supraorbital contrasts with the condition in osteostracans, in

N.II C

N.II Hyoid articulation A Mandibular arch articulation

N.II D Hyoid articulation

Hyoid articulation Otic-occipital fissure Mandibular arch articulation

N.II Mandibular arch articulations

B N.II E Hyoid articulation

Hyoid articulation Otic-occipital Mandibular arch articulation fissure

Mandibular arch articulations

Figure 6. External neurocranial anatomy in lateral view. A, Norselaspis, an osteostracan (after Janvier, 1981b). B, Macropetalichthys, a petalichthyid placoderm (after Stensiö, 1969; Young, 1980). C, Dicksonosteus, an arthrodire placoderm (after Goujet, 1984a). D, Cladodoides, a chondrichthyan (after Maisey, 2005). E, Mimpiscis, a crown osteichthyan and actinopterygian (after Gardiner, 1984b). Abbreviation: N.II, opening for the optic tract (second cranial nerve). Not drawn to scale.

N.VII pal + hm N.VII pal + hm

jugular vein canal jugular vein canal

CD E

N.VII pal N.VII pal

N.VII hm N.VII pal

jugular vein canal N.VII hm jugular vein canal jugular vein canal N.VII hm

Figure 7. Endocranial cavities of various gnathostomes. A, Benneviaspis, an osteostracan (after Janvier, 1985a). B, Brindabellaspis, a placoderm (after Young, 1980). C, Kujdanowiaspis, an arthrodire placoderm (after Goujet, 1984a). D, Buchanosteus, an arthrodire placoderm (after Young, 1979). E, Cladodoides, a crown gnathostome and chondrichthyan (after Maisey, 2005). A, B, in dorsal view. C, D, E, in ventral view. Abbreviations: N.VII, canal or openings for the facial nerve (seventh cranial nerve); hm, hyomandibular branch; pal, palatine branch. Not drawn to scale. which the orbit is surrounded by a broad lateral tine branch. Canals accommodating this division can expansion of the braincase (Fig. 7). As noted by be seen in the cranial endocasts of a number of Janvier (1996b) and Brazeau (2009), a similar condi- Palaeozoic vertebrates: chondrichthyans (Schaeffer, tion is seen in Macropetalichthys (Stensiö, 1925, 1981; Maisey, 2005, 2007), osteichthyans (Jarvik, 1969) and petalichthyid-like forms such as Brin- 1980; Gardiner, 1984b), and arthrodires (Stensiö, dabellaspis (Young, 1980) amongst presumed man- 1963; Young, 1979; Goujet, 1984a). In these forms, the dibulate gnathostomes. A notable non-placoderm facial nerve divides into its several branches behind example is the early chondrichthyan Doliodus the orbit, deep to the postorbital process (Fig. 7C–E). (Maisey et al., 2009). However, given the fact that This contrasts with the condition in osteostracans these expansions are not seen in any other crown (Janvier, 1981b, 1985a), in which the facial nerve is gnathostome, we consider the condition in Doliodus to undivided as it passes into the orbit and along or be secondarily derived. The phylogenetic interpreta- beneath the orbital ﬂoor (Fig. 7A). Its division, con- tion of petalichthyid-like gnathostomes is less clear, sistent with the placement of the hyoid arch (Fig. 6), and the condition found in such taxa could easily be is outside of the postorbital region. The condition is interpreted as a plesiomorphy shared primitively somewhat different in galeaspids (Gai et al., 2011), with osteostracans (Janvier, 1996b; Brazeau, 2009). but the precise position of the facial nerve division The facial nerve (nerve VII) of vertebrates has is not clear. It does not appear to have been medial to several branches, including a somatic sensory the passage of the jugular vein (as it is in crown hyomandibular branch and a visceral sensory pala- gnathostomes). However, the hyoid arch is positioned

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 796 M. D. BRAZEAU AND M. FRIEDMAN immediately posteroventral to the orbit. Thus, the identification of such a diverticulum of the anterior branching of the facial nerve may be more posteriorly end of the saccular chamber is not dependent on the placed than in osteostracans. presence of both an anterior and horizontal canal. Amongst presumed mandibulate gnathostomes, Thus, the diverticulum can be absent in a comparable the facial nerve of Brindabellaspis (Young, 1980) manner in vertebrates with only two semicircular and Macropetalichthys (Stensiö, 1925, 1969) passes canals. through the orbital floor (Fig. 7B), with the hyoman- The jugular vein passes posteriorly from the orbit dibular branch exiting through the lateral wall of the and laterally along the otic capsules. In crown-group orbit. In Brindabellaspis, the division is within the gnathostomes and living agnathans, this vein is not orbital floor. In Macropetalichthys, it appears to have invested in the neurocranium along the lateral been within the orbit itself. This closely resembles sides of the otic capsules (Fig. 7E). However, in conditions in osteostracans, and contrasts with the osteostracans (Janvier, 1981b, 1985a), galeaspids arrangement found in arthrodires, Romundina, and (Gai et al., 2011), and certain placoderms (Brindabel- all known crown-group gnathostomes, in which laspis, Young, 1980; Jagorina, Stensiö, 1969; partially the division is as described above (Janvier, 1996b; in Macropetalichthys, Stensiö, 1925, 1969; and Brazeau, 2009). We regard division of the facial Buchanosteus, Young, 1979), the jugular vein was nerve deep to the postorbital process as a probable completely invested in the otic sidewall (Fig. 7A, B, synapomorphy uniting a subset of jawed vertebrates. D). In many arthrodire placoderms (Stensiö, 1963, Potentially related to the above characters is 1969; Goujet, 1984a) and in Romundina (Ørvig, the position for the articulation of the hyoid arch 1975), the course of the jugular vein is uninvested on the braincase (Fig. 6). In osteostracans, the hyoid laterally (Figs 6C, 7C), and situated in a deep groove and mandibular arches articulate in the anterior in the otic side wall. We conclude that the invested region of a broad rostral expansion of the braincase condition of the jugular vein represents a transient and cranial shield (Fig. 6A). A similar expansion is condition along the gnathostome stem, with the pres- developed in some placoderms to a varying degree. ence of an exposed jugular vein representing a char- The hyoid arch of galeaspids attaches in a mostly acter uniting crown-group gnathostomes and some sub- or postorbital position. The hyoid arch of stem-group members. Macropetalichthys, Brindabellaspis, and ptyctodont Paired nasal capsules are a feature common to all placoderms attaches to the braincase in a position jawed vertebrates, and thus merit some considera- that is either immediately lateral or anterior to the tion here because modern outgroups lack this orbit (Fig. 6B). This position is anatomically inter- feature (see Janvier, 1993). Notably, osteostracans mediate between the condition seen in osteostracans lack paired nasal openings and have a single, median, and the condition apparent in crown-group gnatho- nasohypophyseal opening in the skull, resembling stomes and arthrodires, in which the hyoid arch modern hagfish and lamprey. Recently, Gai et al. attaches behind the orbit (Fig. 6C–E). Although we (2011) reported on the paired nasal capsules in a retain this character here, we caution that it may be Siurian galeaspid from China. Placing osteostracans directly linked to aspects of orbital morphology and as the sister group of gnathostomes to the exclusion facial nerve orientation. of galeaspids suggests an ambiguous distribution The saccular cavity of the gnathostome skeletal for this character. Either paired nasal capsules labyrinth bears a diverticulum, termed the utricular evolved twice (two steps) or were gained once and recess, where the ampullar chambers of the anterior lost in osteostracans (two steps). Resolving this semicircular canal and the horizontal semicircular dichotomy is not simple because impressions of paired canal join it. This chamber is obvious in chondri- nasal capsules or paired openings are reported chthyans (Schaeffer, 1981; Maisey, 2005, 2007), in pteraspidomorphs (e.g. Janvier & Blieck, 1979; arthrodires (Stensiö, 1963, 1969), Macropetalichthys Gagnier, 1993). Owing to the uncertain optimisation (Stensiö, 1925, 1969), Acanthodes (Davis et al., 2012), of this character, we have chosen to omit it from our and osteichthyans (Jarvik, 1980). There is no evi- list until future work resolves this issue. dence of the utricular recess in either osteostracans (Janvier, 1981b, 1985a) or galeaspids (Gai et al., Visceral arches, including palatoquadrate and 2011), and we regard this as the primitive gnatho- Meckelian element stome condition. The utricular recess was reported The most conspicuous synapomorphy of gnathostomes absent in Brindabellaspis and indistinct in Jagorina is dorsoventrally opposing jaws. This directional by Young (1980: 32). We submit that the utricular qualification is significant, given the presence of a recess could be biologically related to the origin of a lateral bite in living agnathans (Yalden, 1985; Clark horizontal semicircular canal, which is absent in all & Summers, 2007) and conodonts (Purnell, 1994; known jawless vertebrates. However, we note that the Purnell & Donoghue, 1997; Donoghue & Purnell,

1999). The jaws of gnathostomes are further equipped because most placoderms have pelvic girdles, the with dermal ossifications bearing denticles. absence of these structures, even in non-antiarchs Gnathostomes are also united by the presence of five such as Lunaspis, ‘can only be interpreted as second- or fewer gill arches. Exceptions to this pattern include ary loss’. We do not find this persuasive, as it employs hexanchiform neoselachians (six- and seven-gilled the specious argument that the most common char- sharks), in which conditions are clearly secondary acter state within a group must be primitive (Watrous (Maisey, Naylor & Ward, 2004). Jawless fishes gener- & Wheeler, 1981; Farris, 1982). Regardless of the ally have higher numbers of gill arches, although precise optimization of this character, it is clear that counts vary considerably both amongst and within the presence of pelvic fins unites the gnathostome agnathan groups. Hagfishes bear between five and 15 crown group and its nearest sister taxa to the exclu- gill arches, with some species showing variation in sion of agnathans. number, whereas lampreys have seven arches. Counts in extinct agnathans are less well constrained, and Summary can often be estimated only for a few exemplars. In The monophyly of jawed vertebrates is supported by his review of gill-arch structure in vertebrates, the following hard-tissue characters. Dermal denticle Janvier (2004) reported the following values: 11 to 20 plates on mandibular cartilages are discussed in a for arandaspids, eight for astraspids, six to 15 for later section along with teeth. Many of these features anaspids, seven to nine for thelodonts, five to 45 for would require homoplasy if the placoderms are galeaspids, and eight to ten for osteostracans. treated as a clade (marked with a superscript ‘P’). 7. Dorsoventrally opposing jaws (Fig. 9). Fins 8. Dermal denticle plates on mandibular cartilages In addition to pectoral appendages, a more posteriorly (Fig. 9). placed set of paired appendages, pelvic fins, are 9. Pelvic fins. known only in jawed vertebrates. Earlier authors 10P. Postorbital process (Figs 7C–E, 8). (Johanson, 2002; Friedman, 2007a; Brazeau, 2009) 11P. Deep branching of nerve VII (Fig. 7C–E). have emphasized the absence of pelvic fins as exclud- 12P. Articulation of hyoid arch with neurocranium ing antiarchs from closer relationships between other posterior to orbit (Fig. 6) placoderms and the gnathostome crown. Zhu et al. 13P. Utricular recess. (2012b) subsequently reported a pelvic girdle in the 14. Horizontal semicircular canal. Early Devonian antiarch Parayunnanolepis, suggest- 15P. Uninvested jugular canal lateral to otic capsule ing that this structure may have been primitively (i.e. discrete lateral commissure; Fig. 7C–E). present in the clade. Young (2010: 538) argued that 16. Five or fewer branchial arches.

Precerebral fontanelle Rhinocapsular division

Anterior dorsal A B fontanelle C D

Posterior Posterior dorsal dorsal fontanelle fontanelle

Otic-occipital fissure

Figure 8. Gnathostome neurocrania in dorsal view. A, Dicksonosteus, an arthrodire placoderm (after Goujet, 1984a). B, Lawrenciella, a crown osteichthyan and crown actinopterygian (after Hamel & Poplin, 2008). C, cf. Cobelodus,a chondrichthyan and possible stem holocephalan (after Maisey, 2007). D, Orthacanthus, a chondrichthyan and possible stem elasmobranch (after Schaeffer, 1981).

CANDIDATE SYNAPOMORPHIES OF PLACODERMS 17. Perichondral division of orbitonasal unit Two characters are unique to placoderms, irrespective (Fig. 8A). of the relationships of other gnathostomes, and the 18. Polarized, matrix-bound dentine cell spaces osteostracan/galeaspid outgroup. (semidentine). Neurocranium In nearly all placoderms for which there is well- CHARACTERS THAT CAN PLACE A FOSSIL IN THE preserved neurocranial material, there is a discrete GNATHOSTOME TOTAL GROUP, CROWNWARD division between the rhinocapsular ossification and OF ANY PLACODERM the rest of the braincase (Fig. 8A; Stensiö, 1963, 1969; Characters in this section concern traits found in Denison, 1978; Goujet, 1984a; Young, 1986). This both osteichthyans and chondrichthyans and which contrasts with chondrichthyans (Schaeffer, 1981; have logical alternatives in jawless vertebrates and Maisey, 2005, 2007; Maisey et al., 2009), osteichthyans placoderms. They therefore imply placement of a (Jarvik, 1980; Gardiner, 1984b), osteostracans fossil crownward of any known mandibulate stem (Janvier, 1985a, b), and galeaspids (Gai et al., 2011), gnathostome. in which such a division is absent. Interestingly, this character has been dismissed as a placoderm synapomorphy in favour of being a generalized verte- Neurocranium brate character (Goujet, 2001). This argument is based Several features of the braincase unite Osteichthyes on the separate chondrification of the nasal capsules in and Chondrichthyes to the exclusion of placoderms vertebrate embryos. However, in neither galeaspids, and agnathans. Most of these have been covered in osteostracans, nor crown-group gnathostomes does detail elsewhere (e.g. Schaeffer, 1981; Young, 1986: such a discrete division persist in adult forms, and this 50; Goujet, 2001: 213; Goujet & Young, 2004). persistence may itself be apomorphic. Nevertheless, Placoderms, along with lampreys and osteostracans the nasal capsules of many crown gnathostomes, such (Janvier, 1975), have an extraocular muscle inner- as Acanthodes (Miles, 1973b; Davis et al., 2012) and vated by cranial nerve IV (trochlear) and which issues chondrichthyans, may be unmineralized or incom- from a myodome posterior to the orbit (Young, 2008). pletely mineralized (Schaeffer, 1981). Thus, separate Called the posterior oblique in these groups, this mineralization may be more generalized (Friedman, muscle appears to be the homologue of the trochlear- 2007a; Brazeau, 2009). However, we recommend that innervated superior oblique in crown-group gnatho- the specific placoderm condition be reinstated to stomes. In living jawed vertebrates, this muscle future analyses of mandibulate gnathostomes because inserts anterodorsal to the foramen for cranial nerve it is clear that the discrete, perichondrally lined divi- II (optic), a condition which we regard as derived sion is absent in other non-placoderm gnathostomes, (Young, 2008). osteostracans, and galeaspids. Placoderms (Stensiö, 1963, 1969; Young, 1980; Goujet, 1984a), osteostracans (Janvier, 1985a, b), Dermal skeleton galeaspids (Gai et al., 2011), and living agnathans We retain semidentine as a possible synapomorphy of share a condition in which the occipital and otic Placodermi with some reservations. The presence of regions of the braincase are co-ossified. Osteichthyans polarized, matrix-bound cell spaces in the dentinous and chondrichthyans are united by the presence tissues of placoderm dermal bones could be interpreted of an oticoccipital fissure (Figs 6D, E, 8), which as intermediate in type between mesodentine, in which is apparent in early members of both groups the cells are matrix bound but not polarized, and (osteichthyans: Jarvik, 1980; Gardiner, 1984b; Yu, (ortho)dentine in which the cells are contained in a 1998; chondrichthyans: Schaeffer, 1981; Maisey, 2005; pulp cavity but have polarized canaliculi (cf. Ørvig, Maisey et al., 2009). This same condition character- 1967; but see Sire et al., 2009). We note that work in izes Acanthodes, which has most recently been progress on placoderm histology has found that char- regarded as a total-group osteichthyan (Miles, 1973b; acterization of semidentine in many taxa is problem- Friedman, 2007a; Brazeau, 2009; Friedman & atic, and interpretations of this tissue type can vary Brazeau, 2010; Davis et al., 2012). These same taxa depending on the orientation of histological sections also bear a posterior dorsal fontanelle (Fig. 8), an and the methods used to visualize them (S. Giles, pers. unmineralized region on the posterodorsal surface of comm., 2013; see also Giles et al., 2013: 639). the braincase located dorsal to the foramen magnum, Summary and which is intersected by the oticoccipital fissure Two characters could most parsimoniously be inter- (Acanthodes: Miles, 1973b; osteichthyans: Gardiner, preted as placoderm synapomorphies under our 1984b; chondrichthyans: Schaeffer, 1981). No such explicit, outgroup-based approach: structure is present in placoderms or any agnathan,

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 EARLY GNATHOSTOME CHARACTERS 799 and might be related to the absence of a persistent the placement of Acanthodes on the chondrichthyan fissure between the occipital arch and otic region. stem, implying that the fissure is a crown- For many decades, a persistent transverse fissure gnathostome symplesiomorphy. The braincase of (termed the ventral cranial fissure; Fig. 9) separating Ptomacanthus is preserved as two sets of minerali- the basisphenoid from the basioccipital was known zations: a basisphenoid region and parachordal only in fish-like osteichthyans and Acanthodes. plates. This taxon has been resolved to either the Under outgroup-based cladistic investigation simi- chondrichthyan stem (Brazeau, 2009) or the gnatho- lar to the one presented here (Miles, 1973b), the stome stem (Brazeau, 2009; Davis et al., 2012). ventral cranial fissure was listed amongst several However, it is unclear whether the separate synapomorphies of Acanthodes and crown-group mineralizations in Ptomacanthus are anatomical or a Osteichthyes. The discovery of such a fissure in two taphonomic artefact. Devonian chondrichthyans (Pucapampella, Janvier & Placoderms (Stensiö, 1963, 1969; Goujet, 1984a), Suarez-Riglos, 1986; Gagnier et al., 1989; Maisey, like osteostracans (Janvier, 1985a) and galeaspids 2001; and an unnamed form from South Africa, (Gai et al., 2011), have braincases that can be Maisey & Anderson, 2001) has challenged this called ‘thick walled’. In each of these groups, the interpretation. However, based on our assumed back- perichondrally mineralized cavum cranii is widely bone tree the character distribution is ambiguous: it separated from the external, mineralized surface of is not possible to distinguish between scenarios the braincase, with this gap indicating the presence of entailing the gain of a cranial fissure and its loss in thick cartilaginous matrix (Fig. 7A–D; Zalc, Goujet & crown chondrichthyans, or the independent acquisi- Colman, 2008). By contrast, the cavum cranii is sepa- tion of the cranial fissure in Pucapampella and rated from the external surface of the neurocranium Osteichthyes. The ambiguity could be resolved by by relatively thin walls in both chondrichthyans

Extended A telencephalonic BC region

Ventral cranial fissure

Ventral Extended cranial fissure D EFtelencephalonic region

Dorsal aortic canals

Figure 9. Gnathostome neurocrania in ventral view. A, Acanthodes, an acanthodian (after Davis et al., 2012). B, Gogonasus, a crown osteichthyan and crown sarcopterygian (after Long, Barwick & Campbell, 1997). C, Mimipiscis, a crown osteichthyan and actinopterygian (after Gardiner, 1984b). D, Dicksonosteus, an arthrodire placoderm (after Goujet, 1984a). E, Cladodoides, a chondrichthyan and possible stem elasmobranch (after Maisey, 2005). F, Pucapampella sp. a probable stem chondrichthyan (after Maisey, 2001).

(Fig. 7E; Schaeffer, 1981; Maisey, 2005, 2007) and tion owing to their greatly elongated olfactory tracts osteichthyans (Poplin, 1974; Jarvik, 1980; Gardiner, (Stensiö, 1925). Conditions found in placoderms cor- 1984b). respond most closely to those in galeaspids (Gai et al., All jawed vertebrates are characterized by the pres- 2011) and osteostracans (Janvier, 1985a), suggesting ence of three semicircular canals (see above), but this is the primitive geometry, whereas osteichthyans Davis et al. (2012, supplement) noted that the junc- and chondrichthyans are united by a relatively elon- tion between the anterior and posterior canals of the gated sphenoid (Brazeau, 2009). bony labyrinth is represented by a well-developed sinus superior only in chondrichthyans (but see Dermal skeleton Orthacanthus; Schaeffer, 1981: fig. 14), osteichthyans, The nature of ‘true’ teeth has been a subject of con- and Acanthodes. By contrast, the union of the siderable debate. There is general agreement that anterior and posterior in all described placoderm structures that may be described as teeth are present endocasts is indistinguishable from their intersection in both chondrichthyans and osteichthyans. Contro- with the saccular chamber (Stensiö, 1969; Young, versy centres on the naming of features present on 1980). The condition found in placoderms agrees with the dermal jaw bones of placoderms (Smith & that in osteostracans (Janvier, 1985a), indicating Johanson, 2003; Young, 2003; Johanson & Smith, that the absence of a sinus superior is a retained 2005; Rücklin et al., 2012). These debates have been plesiomorphy whereas its presence represents a clouded by the application of developmental-process synapomorphy of crown-group gnathostomes. We note definitions of teeth (e.g. Reif, 1982) that are unobserv- that an apparent sinus superior is present in the able in palaeontological material (e.g. the dental galeaspid Shuyu (Gai et al., 2011: supplementary lamina), and by transformational scenarios that lack video 1). However, we consider this nonhomologous any phylogenetic reference (e.g. ‘outside-in’ vs. ‘inside- with that in extant jawed vertebrates based on the out’ hypothesis). The dental lamina – an epidermal assumption that galeaspids are the sister group of invagination in which teeth are preformed – has osteostracans plus all mandibulate gnathostomes. featured prominently in debates about fossils. As Davis et al. (2012: supplement) discussed an addi- it is a soft tissue structure, its presence can only tional feature of the otic capsule that appears to be be inferred through an interpretation of static synapomorphic for crown gnathostomes: the absence of osteological characters in fossils. The choice to inter- a skeletal capsular wall that separates the labyrinth pret some of these characters as necessary and suffi- cavity from the cavum cranii. Such a division cient evidence for a dental lamina – and therefore is absent in early chondrichthyans (e.g. Orthacan- teeth – is not germane to systematic debate as it does thus: Schaeffer, 1981: fig. 14; Cladodoides: Maisey, not constitute independent evidence. 2005: fig. 7), osteichthyans (e.g. Mimipiscis: Gardiner, We are primarily interested in the systematic 1984b: fig. 26; Psarolepis: Yu, 1998: fig. 4), and Acan- relationships of problematic fossil taxa, and not with thodes (Davis et al., 2012: supplementary fig. 10). By the definition of the word ‘tooth’. We have therefore contrast, the labyrinth cavity is largely divided from resolved teeth into hierarchical, conditional state- the remainder of the endocranial cavity in osteo- ments of homology. This reflects possible relation- stracans (Janvier, 1985a, b), galeaspids (Gai et al., ships at the levels of oral dermal tubercles alone, 2011), and a variety of placoderms (e.g. Kujdano- or variations on this condition including precise wiaspis, Tapinosteus, Macropetalichthys, Jagorina: ordering/patterning, replacement, topology, or other Stensiö, 1969: figs 44, 46, 48, 52; Brindabellaspis: conditions. Young, 1980: fig. 10; Dicksonosteus: Goujet, 1984a: Amongst placoderm-grade fishes, some bear tuber- fig. 26; possibly Bolivosteus, Goujet, Janvier & cles along their dermal jaw bones whereas others, so Suarez-Riglos, 1985: fig. 3), indicating that the condi- far as can be determined, do not. Regardless of any tion apparent in crown gnathostomes is derived. Davis additional criteria, these tuberculations are accept- et al. (2012: supplement) argued that a capsular wall is able primary homologues of teeth insofar as they are absent in ptyctodonts. However, we regard the condi- kinds of dermal tubercle borne on the biting surfaces tion in this group as unclear, given uncertainties of the jaws; this holds even if they do not satisfy concerning neurocranial structure arising from the essentialistic definitions of teeth. Such oral denticles presence of multiple ossification centres. may unite some placoderms with the gnathostome The sphenoid region and telencephalon of placo- crown to the exclusion of others, support jawed ver- derms (Stensiö, 1963, 1969; Young, 1980; Goujet, tebrate monophyly with losses in some placoderm 1984a) are comparatively much shorter than those of groups, or represent unique derivations within par- chondrichthyans (Schaeffer, 1981; Maisey, 2005, 2007) ticular placoderm clades. and osteichthyans (Jarvik, 1980; Gardiner, 1984b), It is only in a subset of arthrodire placoderms that with petalichthyids representing an important excep- these tubercles are manifest in a precisely patterned

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 EARLY GNATHOSTOME CHARACTERS 801 arrangement that resembles that of the teeth of (e.g. ischnacanthids: Watson, 1937; Climatius: Miles, crown-group gnathostomes (Smith & Johanson, 2003; 1973a; ‘Protodus’: Burrow & Turner, 2010; Ptoma- Johanson & Smith, 2005; Rücklin et al., 2012). The canthus: Brazeau, 2012; Latviacanthus: Schultze & question of whether or not the precise patterning of Zidek, 1982), sarcopterygians (onychodonts: Andrews these structures is homologous to that apparent et al., 2005; porolepiforms: Jarvik, 1972), we consider in the teeth of crown gnathostomes pivots on the this to be the general pattern for crown-group systematic status of placoderms generally and gnathostome teeth. Contrasts are provided by most arthrodires specifically. However, there seems to be no actinopterygians (Pearson & Westoll, 1979; Gardiner, reason to regard arthrodire teeth (indeed, those of 1984b; but see Howqualepis, in which anterior many placoderms) and those of crown gnathostomes dentary teeth are arrayed in a whorl-like fashion: as nonhomologous insofar as they are both kinds of Long, 1988) and some probable stem osteichthyans dental tubercles forming biting plates on the man- (e.g. Dialipina; Schultze & Cumbaa, 2001), and we dibular arch, a feature optimized to the last common caution that our inference may be extremely sensitive ancestor of both groups regardless of whether to new osteichthyan data. Dentitions of crownward placoderms form a clade (e.g. Smith & Johanson, members of the osteichthyan stem remain effectively 2003; Johanson & Smith, 2005; Rücklin et al., 2012) undocumented, following re-interpretation of putative or grade (Friedman, 2007a; Brazeau, 2009; Davis teeth in the stem osteichthyan Andreolepis and et al., 2012). incertae sedis taxon Lophosteus (Cunningham et al., Given the broad distribution of whorl-like cusp files 2012). (Fig. 10) in chondrichthyans (Zangerl, 1981; Williams, The recent discovery of Entelognathus (Zhu et al., 2001; Maisey et al., 2009), many acanthodians 2013) a placoderm-grade fish with marginal jaw bones

Figure 10. Gnathostome dental anatomy. A, Torosteus pulchellus, NHMUK P.50966, an arthrodire placoderm. B, Ischnacanthus gracilis, NMS 1887.35.2, an acanthodian. C, Cladoselache sp., NHMUK P.9272, a crown gnathostome and chondrichthyan. D, Onychodus jandemarrai, NHMUK P.63576 (image reversed), a crown osteichthyan and sarcopterygian. Scale bars = 10 mm.

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 802 M. D. BRAZEAU AND M. FRIEDMAN and gular armour has raised questions about the giving rise to an ‘onion skin’ appearance when viewed phylogenetic distribution of these traits. In discussing in histological cross-section. This unique mode of osteichthyan characters, Friedman & Brazeau (2010) development does appear to characterize a subset of included ‘differentiated branchiostegals (opercular, spine-bearing early gnathostomes, in particular the subopercular, lateral gulars)’ in their list of osteich- acanthodiforms and diplacanthids. However, the term thyan synapomorphies. Their presence or absence ‘Acanthodii’ is more broadly applied, and includes was not considered contingent on the presence of taxa such as Parexus, Climatius, and Ptomacanthus, more general macromery of the dermal skull. This is which have been reported to have areally growing because branchiostegal plates are common in many scales (Burrow & Turner, 2010; Brazeau, 2012; acanthodian-type fishes that have micromeric skulls. Burrow et al., 2013). Ischnacanthids and Brocho- In addition to being absent in chondrichthyans and admones have unusual conditions in which odon- all fishes assigned to the Acanthodii, they were absent tode generations do not cover the entire crown in any placoderm-grade fishes prior to the discovery of (Valiukevicius, 1992; Hanke & Wilson, 2006). Owing Entelognathus. to the patchy distribution of this character, we are Assuming that Entelognathus is a stem gnatho- omitting it from our list, but we accept that it prob- stome (based on the apparent absence of cranial ably identifies a monophyletic subset of taxa com- fissures, the absence of teeth, and lack of any monly referred to as acanthodians. specific apomorphies of either chondrichthyans or osteichthyans – except potentially the facial jaw Fins bones and gulars), we envision two possible scenarios The presence of dermal projections associated with for the origin of facial jaw bones and gular plates. the pelvic girdle in placoderms (Long & Young, 1988), Either they were two independent gains or one gain stem-group sarcopterygians (Zhu et al., 2012a), and one loss. Owing to this uncertainty, we follow Zhu and possibly early chondrichthyans (Miller et al., et al. (2013) in not placing Entelognathus in a 2003: 503) suggests that pelvic spines are not resolved position with respect to arthrodires and the a synapomorphy uniting acanthodians. However, gnathostome crown. acanthodians do appear unique amongst vertebrates in bearing a well-developed spine in association Summary with the anal fin (Fig. 4D; Maisey, 1986). This struc- The following are derived features that can place a ture, which is absent in all known placoderms, fossil crownward of placoderms. chondrichthyans, and osteichthyans, is present 19. Extrinsic eye muscle innervated by the trochlear in all classically recognized acanthodian orders nerve (N.IV) originates anterodorsal to optic (Denison, 1979). An anal fin spine is retained in the foramen. acanthodian Paucicanthus, which lacks spines in 20. Otico-occiptial fissure. association with its paired fins (Hanke, 2002). A 21. Posterior dorsal fontanelle (Fig. 8B–D). series of more complete body fossils bearing anal-fin 22. Ventral fissure (Fig. 9). spines has recently been attributed to either the 23. Thin endocranial wall. Chonrichthyes or Teleostomi incertae sedis (Hanke & 24. Sinus superior. Wilson, 2004, 2010; Hanke et al., 2013) on the basis of 25. Labyrinth cavity confluent with remainder of scale morphology. This implies several competing sce- endocranial chamber (capsular wall absent). narios for anal fin spine distributions, which are 26. Extended telencephalonic region (and/or corre- discussed later in the text. spondingly elongate sphenoid). 27. Teeth (possibly homoplasious in arthrodires; Summary Fig. 10). We recognize a single character as a potential synapomorphy of acanthodians: CANDIDATE SYNAPOMORPHIES OF ACANTHODIANS We retain here all characters common to classically 28. Anal-fin spine present (Fig. 4D). defined acanthodians. We exclude the proposed acanthodian synapomorphy concerning the shoulder CHARACTERS THAT CAN PLACE A FOSSIL girdle (Burrow & Turner, 2010; Hanke & Davis, 2012) IN THE GNATHOSTOME CROWN GROUP: for reasons given above in our discussion of compound CHONDRICHTHYAN SYNAPOMORPHIES characters. Positively placing a taxon within the gnathostome Dermal skeleton crown requires uniting it with either Chondrichthyes Acanthodians have been characterized as possessing or Osteichthyes. We provide an account of chond- scales with concentrically apposed generations – richthyan characters here, along with an updated

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 EARLY GNATHOSTOME CHARACTERS 803 synapomorphy scheme for osteichthyans from Amongst fossil chondrichthyans, many taxa exhibit Friedman & Brazeau (2010) in the next section. a pair of canals for the lateral dorsal aortae within In spite of broad consensus on the monophyly of the basioccipital cartilage. This includes Doliodus Chondrichthyes, highly inclusive total-group charac- and Pucapampella (Maisey, 2001; Maisey et al., ters for chondrichthyans have been elusive and this 2009), taxa that are reasonably interpreted as stem group has generally been identified by the absence of chondrichthyans, as well as forms that are routinely osteichthyan features (Maisey, 1986: 216). Tessellate interpreted as members of the chondrichthyan crown prismatic calcified cartilage (Fig. 5C, D) seems to be group (Schaeffer, 1981; Coates & Sequeira, 1998). the only undisputed chondrichthyan synapomorphy. This contrasts with the condition seen in arthrodires Many of the characters detailed below may well be and osteichthyans. A canal for the medial dorsal elasmobranch synapomorphies, rather than more aorta is seen in early actinopterygians, but the general chondrichthyan traits, but present evi- divided portion of the artery is uninvested in cartilage dence does not allow an unequivocal conclusion. We (Gardiner, 1984b). This supports an interpretation of are concerned here with total-group membership this feature as either an elasmobranch synapomorphy rather than discriminating between stem and crown (Coates & Sequeira, 1998, 2001) or a chondrichthyan chondrichthyans; any feature that is a synapomorphy crown-group synapomorphy (Maisey, 2001). The of Elasmobranchii (all crown-group chondrichthyans absence of canals for the lateral dorsal aortae is not that are not members of total-group Holocephali) universal in placoderms, however. Paired canals are must also be evidence of membership within the found in Brindabellaspis, and modest canals are chondrichthyan total group. This is an admittedly seen in Macropetalichthys (see Young, 1980). We unsatisfactory situation, but the uncertain status of retain this character as evidence of chondrichthyan the characters reviewed below can be taken to indi- total-group membership, but note that the status of cate the current state of early chondrichthyan this character hinges on the phylogenetic interpreta- systematics. The discussion of potential chondrich- tion of placoderms. thyan synapomorphies presented here is maximally The hypotic lamina consists of a broad lateral inclusive, in the sense that it reviews those features extension of the parachordal plate that underlaps the supporting chondrichthyan monophyly if all acantho- otic capsules. This feature is present in a number of dians can be assigned to the total group. Such Palaeozoic chondrichthyans in which the relationship features, which are either gnathostome symplesio- between the parachordal plate and the otic capsules morphies or of equivocal polarity when some or all can be inspected (Schaeffer, 1981; Maisey, 2005, acanthodians are placed on the osteichthyan stem, 2007). Amongst chondrichthyans that lack a distinct are indicated explicitly below and in the following metotic fissure, the presence or absence of the trait character list. can be verified in embryonic stages showing distinct neurocranial divisions (De Beer, 1931). The feature is Neurocranium absent in either adult or embryonic specimens of A broad anterodorsal opening of the cranial cavity Callorhincus (see De Beer & Moy-Thomas, 1935), a between the nasal capsules (Fig. 8) is a remarkably state that may be general to holocephalans. Embry- stable feature of shark-like chondrichthyans and onic information is lacking in fossil taxa, especially batoids from the Palaeozoic to the Recent (Schaeffer, placoderms, making it difficult to assess the polarity 1981; Maisey, 2005, 2007; Maisey et al., 2009). of this character within crown-group gnathostomes. It is absent in adult holocephalans. De Beer & Although there is a clearly observable condition that Moy-Thomas (1935) interpreted the ethmoid canal is restricted to a subset of fossil chondrichthyans, in embryos of Callorhincus as a corresponding struc- we currently exclude this character from our list ture, leading Schaeffer (1981) to propose this feature because of the uncertainty concerning outgroup as a generalized trait of chondrichthyans. Maisey conditions. Nevertheless, Maisey (2001) reported (2005: 288) retained the stricter interpretation of the hypotic lamina as absent in the probable stem the precerebral fontanelle as an elasmobranch chondrichthyan Pucapampella. It is therefore possible synapomorphy, because of differences in interpre- that this character is restricted to a more limited tation of the ethmoid canal. However, as noted subset of the chondrichthyan crown or to total-group by Maisey (2005), the hypothesis that some elasmobranchs. Resolution of this will be greatly elasmobranch-like Palaeozoic chondrichthyans are assisted by improved documentation of the anatomy stem holocephalans (e.g. Coates & Sequeira, 2001; of Pucapampella. Pradel et al., 2011), as well as its presence in Doliodus A dorsal otic ridge has been proposed as a (a probable stem-group chondrichthyan) invites the chondrichthyan synapomorphy (Maisey, 2001; this is possibility that the absence of a precerebral fontanelle mentioned, but not listed as a chondrichthyan in holocephalans is secondary. synapomorphy, by Schaeffer, 1981). This consists of a

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 804 M. D. BRAZEAU AND M. FRIEDMAN distinct, anteroposteriorly orientated, midline ridge Dermal skeleton situated in front of the endolymphatic fontanelle on Most Palaeozoic chondrichthyans lack any signifi- the otic portion of the braincase (Fig. 8C, D). The cant dermal skull plates (Zangerl, 1981; Stahl, 1999). ridge is flanked on either side by a pair of depres- A notable exception is found in some early sions, which were probably insertion areas for the holocephalans (Patterson, 1965; Stahl, 1999), but we epaxial muscles. A comparable ridge has not been consider this probably a derived state. Outgroup observed in any placoderm (Stensiö, 1969), and is comparison with agnathans is equivocal, but osteo- absent in Devonian osteichthyans (Gardiner, 1984b). stracan and galeaspid skull roofs appear to have been The ridge was recently identified in Acanthodes by primitively composed of individual tesserae (Janvier, Davis et al. (2012). The placement of Acanthodes as a 1996a; Sansom, 2009). Placoderm and osteichthyan stem osteichthyan in recently published cladograms skull roofs, however, are generally macromeric, (Brazeau, 2009; Davis et al., 2012) contradicts the although tessellate conditions appear in certain interpretation of this feature as a synapomorphy placoderms. Under our outgrouping scheme, the of chondrichthyans, implying instead that it is a status of micromery as a chondrichthyan apomorphy gnathostome symplesiomorphy. However, if we apply is ambiguous. We must assume at least two independ- the logic used throughout this paper, then the absence ent acquisitions of macromery (in osteichthyans and of this feature in all outgroups (osteichthyans, all placoderms), or one gain of macromery and a subse- placoderms, and proximal jawless groups), then we quent reversal to micromery (in chondrichthyans must count this character as evidence against the and acanthodians). The resolution between these osteichthyan placement of Acanthodes. alternatives further depends upon the placement What is problematic about this character is its of acanthodians relative to the two crown gnatho- potential irrelevance to taxa with strongly attached stome lineages. If the members of this assemblage skull roofing bones. Although the character is absent branch from the osteichthyan, chondrichthyan, and in chondrichthyan outgroups considered here, it gnathostome stems (Brazeau, 2009; Davis et al., appears in the actinopterygian Lawrenciella (see 2012), then micromery would map as the generalized Fig. 8B; Hamel & Poplin, 2008), a taxon only known condition for crown gnathostomes. However, if from isolated braincases for which associated roofing acanthodians are restricted to the chondrichthyan bones have not been recovered. The arrangement stem, then micromery would probably represent a in Lawrenciella is very probably convergent on the synapomorphy of a subset of the chondrichthyan total chondrichthyan condition, but that is itself poten- group. tially significant because the presence of a ridge is Isolated scales and teeth are the most commonly associated with taxa without firmly attached skull found fossil remains assigned to the Chondrichthyes. roofing bones. We therefore caution that the absence In spite of a dearth of articulated body fossils of of this feature in taxa with and without a tight Devonian and Silurian chondrichthyans, isolated association between the dermal skull roof and the material is routinely attributed to the group without neurocranium might not reflect the same state. contest. Scale-based taxa have thus seized the reins Nevertheless, it is absent in some chondrichthyan of debate to the degree that articulated specimens groups. If it is not a chondrichthyan synapomorphy, it bearing superficially ‘shark-like’ scales, but otherwise might be informative for chondrichthyan ingroup lacking any obvious synapomorphies of either crown relationships. gnathostomes or jawed vertebrates more broadly, have been identified as chondrichthyans (Märss, Wilson & Thorsteinsson, 2002). Such interpretations Primary skeleton are combined with the persistent identification of very The endoskeleton of chondrichthyans comprises one early chondrichthyans based on ‘shark-like’ scales or more layers of tessellate prismatic calcified carti- that cite overall similarity rather than distinct char- lage (Fig. 5C, D; Dean & Summers, 2006). The sig- acters as evidence (e.g. Sansom et al., 2012: 246). nificance of this character as a chondrichthyan These conclusions are symptomatic of the hazy cri- synapomorphy has been well established and corro- teria for identifying early chondrichthyans on the borated by phylogenetic analysis (Goodrich, 1909; basis of skeletal debris, and have the potential to Schaeffer, 1981; Maisey, 1986; Brazeau, 2009; substantially distort our understanding of the timing Davis et al., 2012). Studies of living and fossil of evolutionary divergences within gnathostomes and chondrichthyans have established its ubiquity in vertebrates more generally. We regard these phenom- elasmobranchs, holocephalans, and the various ena as especially problematic, given that the earliest Palaeozoic taxa within this dichotomy (e.g. Kemp & body fossils that can be identified as chondrichthyans Westrin, 1979; Schaeffer, 1981; Stahl, 1999; Dean & on the basis of unambiguous synapomorphies bear Summers, 2006). scales that have effectively gone unstudied (e.g.

Doliodus, Protacrodus; but see Gladbachus: Burrow Gross, 1968; Märss, 1979, 1986; but see putative & Turner, 2013). cranial sensory lines between scale rows in Wilson & Schemes for the attribution of microremains to Caldwell, 1998: 19), the lateral line canal of the trunk chondrichthyans consist of abstracted descriptions of passes through a perforation or groove in a single scale types, largely derived from isolated scales for scale, rather than between rows of scales. Skeletal which there is no additional anatomical information signatures for sensory lines on the trunk are unknown (Karatajute-Talimaa, 1998). No clear indications in anaspids (Ørvig, 1972). Cephalic sensory lines (i.e. synapomorphic hierarchy) are offered as to why that lie within grooves or canals on the head shields these differing scale types might derive from of pteraspidomorphs (Halstead, 1973; Sansom et al., chondrichthyans. Nevertheless, published accounts 1997), and pore-like openings in the scales of Anglaspis of scales derived from chondrichthyan skeletons heintzi are described by Blieck & Heintz (1983). In clearly show areal growth patterns (e.g. Dick, 1981; Sacabambaspis, the ventral lateral line grooves extend Williams, 1998) – that is, growth comprising com- onto the body scales. plexes of primarily appositionally joined odontodes The extension of the main lateral line between (as opposed to burial and/or resorbtion). We have (rather than within or on) scales was initially not seen evidence of this type of scale in skeletons interpreted as a gnathostome symplesiomorphy of fishes that are demonstrably not chondri- by Friedman & Brazeau (2010) on the basis of chthyans. Nevertheless, Palaeozoic chondrichthyans an assumed osteichthyan affinity for Acanthodes. may exhibit a variety of scale morphologies, ranging However, using our outgroup-based approach here, from placoderm-like scales in Gladbachus (Burrow the signal is clear: this character would be most & Turner, 2013) to monodontode denticles (e.g. parsimoniously interpreted as a synapomorphy Hamiltonichthys; Maisey, 1989). Thus, areal scale uniting all acanthodians with the chondrichthyan growth is apparently unique to chondrichthyans, but total group. We therefore consider this evidence it is uncertain whether it is the most inclusive against an osteichthyan identity for all acanthodians, synapomorphy so far identified. like the scapular blade discussed above, but note that We contend that derived features of early both are incongruent with other apparently derived chondrichthyan dentitions can be more reliably features shared between some acanthodians and isolated at present than those of scales. Based on osteichthyans (see below). their broad distribution amongst chondrichthyans, osteichthyans, and acanthodians, whorl-like tooth Fins files would appear to be a general feature of crown The gnathostome pectoral girdle comprises an gnathostomes. We consider individualized tooth bases endoskeletal component, the scapulocoracoid, which in such whorls to be the specialized chondrichthyan supports the pectoral fin skeleton and serves as an trait based on its presence within elasmobranchs and insertion for fin muscles. In chondrichthyans, the some early holocephalans (e.g. Helodus; Moy-Thomas, scapulocoracoid bears a tall blade or process that 1936). It is notable that the bases of teeth assigned to forms the posterior boundary of the gill chamber and Doliodus problematicus exhibit a thin lamina of bone provides an origin for the neck muscles and anterior joining them (Turner, 2004). This is consistent with a trunk muscles (Zangerl, 1981; Liem & Summers, placement of Doliodus as the sister group of most 1999; Stahl, 1999; Gudo & Homberger, 2002; Coates other chondrichthyans (Brazeau, 2009; Davis et al., & Gess, 2007; Anderson, 2008). A similar process 2012). It also suggests that individualized bases is observed in a large number of acanthodians evolved somewhat later than the precerebral fonta- (Miles, 1973a; Denison, 1979), and is only absent in nelle and a more robustly mineralized endoskeleton, taxa in which the whole pectoral endoskeleton is meaning that this will be of little value in resolving unpreserved. In osteichthyans and placoderms in the acanthodian problem. which the scapulocoracoid is preserved, a pronounced Many fossil chondrichthyans show the lateral line scapular blade is absent (Jarvik, 1980; Gardiner, canal passing between flank scales (Cladoselache, 1984b; Goujet, 1984a). Davis et al. (2012) scored a Ctenacanthus: Dean, 1909; Diademodus: Harris, scapular process as present in ptyctodontids based on 1951; Hamiltonichthys: Maisey, 1989; Gladbachus: the figures in Miles & Young (1977). Although we are Friedman & Brazeau, 2010). This condition is not rejecting characters based on overall similarity, seen in all acanthodians (Ørvig, 1972). In osteich- this projection seems to have hardly anything in thyans (Jarvik, 1980; Gardiner, 1984b), placoderms common with the scapular process of chondrichthyans (see Pterichthyodes, Hemmings, 1978; Sigaspis Goujet, and acanthodians. It appears to be quite removed 1973; Groenlandaspis, Burrow & Turner, 1999: fig. 4i), from any participation in the posterior margin of the and jawless stem gnathostomes (osteostracans: branchial chamber, or any likely origin for the neck Sansom, Rodygin & Donoghue, 2008; thelodonts: muscles (Trinajstic et al., 2012: fig. 4). It may be

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 806 M. D. BRAZEAU AND M. FRIEDMAN homologous as a modest dorsal projection, but then it wider distribution of these structures within would lack the conditions that we have described placoderms than previously realized. The presence above, and which we consider derived states. We also of claspers in ptyctodonts has long been known acknowledge that this character might be correlated (Ørvig, 1960), and these structures have generally with the absence of macromeric shoulder plates, been considered independently derived relative to with the scapular spine assuming the functional place chondrichthyan examples (e.g. Denison, 1978). The of the cleithrum/anterolateral plate or vice versa. absence of pelvic claspers in osteichthyans, as well as However, there is no logical reason why such a dermal all acanthodians (some of which are associated with plate and endoskeletal scapular blade could not the chondrichthyan and gnathostome stems in recent co-exist (as they appear to in a limited way in some analyses; Brazeau, 2009; Davis et al., 2012), suggests acanthodians: Miles, 1973a). Although previous place- that chondrichthyan claspers are not homologous ments of acanthodians on both the chondrichthyan with those found in placoderms on the basis of char- and osteichthyan stems (Brazeau, 2009; Davis et al., acter distribution. 2012) suggest that the presence of a scapular process Dorsal fin spines are recognized in chondrich- is a general feature of crown gnathostomes, this thyans (Zangerl, 1981), acanthodians (Denison, 1979), feature is most parsimoniously interpreted as a osteichthyans (Zhu et al., 1999, 2009), and placoderms synapomorphy of acanthodians and chondrichthyans (Denison, 1978). In most Palaeozoic chondrichthyans under our approach. We therefore regard this feature in which the dorsal spine is preserved, a series as evidence against an osteichthyan identity for some of denticles can be observed on the posterior acanthodians. or posterolateral margins (Fig. 11; Zangerl, 1981; Several additional median- and paired-fin charac- Brazeau, 2009). This includes early holocephalans ters as synapomorphies of chondrichthyans (Maisey, (Patterson, 1965; Stahl, 1999). Such denticles are also 1986). Some of these are widely distributed (e.g. present on the dorsal fin spines of some acanthodians basal cartilages of the dorsal fin, which are also found such as Brochoadmones (Fig. 11; Hanke & Wilson, in osteichthyans, acanthodians, and placoderms; 2006) and Parexus (Watson, 1937). Their precise Watson, 1937; Ørvig, 1960; Friedman & Brazeau, location can be somewhat varied, being either at 2010), or do not appear to be characteristic of all the margins of the posterior face of the spine or or even most early members of the group (e.g. more medially aligned along this posterior surface. premetapterygial ‘basals’ present, with each articulat- These variable placements might be individually ing with multiple radials; Zangerl, 1981). From this apomorphous conditions, rather than suggesting list, we retain the presence of pelvic claspers as nonhomology of these denticles. We note however that a chondrichthyan synapomorphy. Our assessment the distribution of such denticles is more sporadic in is complicated by the presence of claspers in paired fin spines. We therefore emphasize that this ptyctodonts, as well as their recent discovery in character only supports chondrichthyan affinity when arthrodires (Ahlberg et al., 2009), which indicate referring to dorsal fin spines.

Figure 11. Dorsal ﬁn spines with trailing edge denticles, a potential synapomorphy of chondrichthyans. A, Brochoadmones milesi, UALVP 41495, an acanthodian. B, Tristychius arcuatus, NHMUK P.11378-79, a crown chondrichthyan and stem elasmobranch. Scale bars = 10 mm.

Summary The ascending basisphenoid pillar pierced by the The following characters represent derived features common internal carotid is conditional on the pres- that permit the placement of a fossil in the ence of an anatomical structure comparable to a chondrichthyan total group. Characters marked with basisphenoid pillar. This is absent in platybasic taxa a superscript ‘A’ have potentially ambiguous distri- (which includes most non-osteichthyans). This makes butions based on the placement of Acanthodes and this a highly contingent character that could, at best, its relations (e.g. other acanthodids such as identify only more crownward members of the Homalacanthus and Cheiracanthus, and probably also osteichthyan stem, but by itself is not evidence inde- Mesacanthus and Promesacanthus): pendent of the presence of a dorsally restricted endocranial cavity in the braincase of osteichthyans. 29. Precerebral fontanelle (Fig. 8C, D). We eliminate this character in order to maintain 30. Canals for lateral dorsal aortae in basioccipital. consistency across the present contribution. A 31 . Dorsal otic ridge (Fig. 8C, D). The presence of dorsal scutes has been regarded as 32. Tessellate prismatic calcified cartilage (Fig. 5C, a derived character of either osteichthyans (Friedman D). & Brazeau, 2010) or a subset of that group (Patterson, A 33 . Micromeric skull. 1982b). However, ridge scales of some description 34. Scales with areally apposed odontodes. are present in a wide variety of total-group gnatho- 35. Tooth files with unfused bases (Fig. 10C). stomes, including acanthodians (e.g. Brachyacanthus: 36. Denticles on fin side of dorsal-fin spines (Fig. 11). Watson, 1937), placoderms (e.g. Lunaspis: Gross, 37. Pelvic claspers. 1961; Stensioella, Paraplesiobatis: Gross, 1962; A 38 . Dorsal scapular blade. Sigaspis: Goujet, 1973; Parayunnanolepis: Zhang, A 39 . Lateral sensory line passing between flank Wang & Wang, 2001), and many osteostracans scales. (Janvier, 1985b, 1996a). In light of this distribution and ambiguities surrounding the placement of acanthodians (see below), we no longer consider the CHARACTERS THAT CAN PLACE A FOSSIL IN presence of dorsal ridge scales a reliable osteichthyan THE GNATHOSTOME CROWN GROUP: synapomorphy. Similarly, the presence of gular plates OSTEICHTHYAN SYNAPOMORPHIES in the probable stem gnathostome Entelognathus Characters supporting the monophyly of Osteichthyes (Zhu et al., 2013) casts doubt on prior claims that have already been discussed in detail by Friedman & such structures are derived features of osteichthyans Brazeau (2010). They will not be treated extensively (Friedman & Brazeau, 2010). here. We have made a few amendments based on As with our list of chondrichthyan synapomorphies, recent publications (Maisey, 2007; Brazeau, 2009; the list presented here is maximally inclusive. Here, Davis et al., 2012; Zhu et al., 2013). our list reflects osteichthyan synapomorphies given Tropibasy has recently been clarified in terms placement of Acanthodes on the osteichthyan stem of adult anatomical conditions by Maisey (2007). (Miles, 1973a; Friedman, 2007b; Brazeau, 2009; Three main anatomical conditions are observed in Friedman & Brazeau, 2010; Davis et al., 2012). osteichthyans: a narrow or septate division between However, those characters that are no longer recon- the orbits, that this septum or division consists of a structed as synapomorphic for Osteichthyes under deep endochondral floor between the cranial cavity at alternative placements of Acanthodes and its immedi- the ventral surface of the basisphenoid, and that it is ate relations are indicated here with a superscript ‘A’. pierced by the common internal carotid. The first two An asterisk indicates that the character becomes of these conditions are observed in a braincase com- ambiguous and unresolvable if Acanthodes is removed parable to Cobelodus (Maisey, 2007). Furthermore, it to the chondrichthyan stem (cf. Zhu et al., 2013), owing is doubtful whether many anatomically primitive to uncertainty in outgroups. Contrary to Zhu et al. osteichthyan braincases (e.g. Ligulalepis, Guiyu, (2013) and Friedman & Brazeau (2013), Friedman & Psarolepis, Onychodus) have a narrow interorbital Brazeau (2010) did not list the premaxilla, maxilla, space. What does appear to be limited to osteich- and dentary bones as osteichthyan synapomorphies thyans (and symmoriforms, which appear to be highly because they are contingent on the presence of clustered within Chondrichthyes; Pradel et al., 2011) macromeric skull roofing bones. is a dorsally restricted endocranial cavity in the sphenoid portion of the braincase – regardless of whether 40A. Two or fewer spino-occipital nerve openings. this region is mediolaterally narrow. We have further 41. Endocranial cavity dorsally restricted within refined the osteichthyan condition to reflect this, sphenoid (i.e. between orbits). rather than the complex character compound implied 42A. Ventral surface of otic capsules mediolaterally in ‘tropibasy’. sloping.

43A. Spiracular grooves on basisphenoid. placoderm skulls, notably the architecture of the 44A. Vestibular fontanelles. braincase of arthrodires plus Romundina relative 45. Horizontal semicircular canal joins level with to some or most other placoderms. Arthrodires the posterior ampulla (Davis et al., 2012). and Romundina differ from osteostracans and some 46. Absence of endolymphatic duct openings in placoderms in the presence of laterally open orbits dermal skull. (i.e. the orbits open on the lateral, rather than dorsal, 47A. Anterior dorsal fontanelle (Fig. 8B). face of the dermal skull; Fig. 6), incompletely invested 48A. Enamel. jugular vein lateral to the otic capsule (Fig. 7C), 49. Endochondral bone (Fig. 5B). laterally open orbits possessing a suborbital shelf 50A. Hyomandibular articulates with lateral (Figs 6, 7), deep branching of the facial nerve (Fig. 7C, commissure/postorbital process. D), and a suborbital attachment of the mandibular 51. Hyomandibular branch of the facial (N.VII) arch (complemented by a postorbital attachment nerve exits into jugular canal. of the hyoid arch; Fig. 6). This suggests that even 52. Long canals for olfactory tracts. our proposed placoderm synapomorphies might also 53. Ethmoid comineralized with sphenoid and com- be considered gnathostome symplesiomorphies, a pletely encloses the nasal capsules. hypothesis we currently favour. 54. Hypohyal linking ceratohyal and basihyal. Arthrodires are quite reasonably interpreted in 55. First two gill arches articulate with common light of crown-group gnathostomes, with which they basibranchial. share many specializations. This is reflected in their 56. Suprapharyngobranchials. frequent deployment as an outgroup in phylogenetic 57A*. Anteriorly directed infrapharyngobranchials. analyses of osteichthyans and chondrichthyans (e.g. 58*. Hyomandibular bears deep groove or canal for Zhu et al., 1999, 2001, 2006, 2009; Coates & Sequeira, hyomandibular branch of facial nerve (possibly 2001; Zhu & Yu, 2002). However, other placoderms independently derived within the crown). are frequently, by extension, interpreted in light 59. Biconcave glenoid on lower jaw. of arthrodires (e.g. Young, 1980) under the assump- 60A. Macromeric dermal skull. tion that placoderms form a natural group to the 61A. Pectoral girdle with substantial dermal exclusion of all other jawed vertebrates. We and component. others (e.g. Janvier, 1996b) have highlighted the deep 62. Lepidotrichia. division in placoderms between those with remark- 63. Proximally restricted radials supporting ably osteostracan-like braincase anatomy, and those, hypochordal lobe of caudal fin. such as arthrodires, that closely resemble crown 64. Dermal bones lining palate and lower jaw (e.g. gnathostomes. As easy as it is to interpret arthrodires entopterygoid, prearticular). in terms of crown gnathostomes, so too is it easy to interpret Brindabellaspis and Macropetalichthys in terms of any given osteostracan or, potentially, IMPLICATIONS FOR CLASSICAL galeaspid. The recent discovery of Entelognathus (Zhu et al., TAXONOMIES OF EARLY VERTEBRATES 2013) highlights a further piece of evidence difficult to PROBLEMS WITH PLACODERM MONOPHYLY reconcile with placoderm monophyly. Entelognathus Our investigation identifies two candidate placoderm is an arthrodire-like fish that exhibits marginal jaw synapomorphies: a separate rhinocapsular ossifica- bones and a facial jaw and gular skeleton, previously tion and the presence of semidentine. Nevertheless, unreported in any placoderm. Placoderms are not we remain circumspect concerning placoderm rare fossils, and a number of species are known monophyly. These two proposed synapomorphies are from articulated specimens. Entelognathus therefore not unproblematic in their interpretation, nor have comes as something quite unexpected (Friedman & they been universally observed in placoderms (see Brazeau, 2013). It provides strong corroboration for a discussions above). The separate rhinocapsular crownward placement of arthrodire-like taxa, but not ossification can only be observed in taxa with miner- without implying that all acanthodians are total- alized neurocrania, although it is taxonomically wide- group chondrichthyans. spread. However, it has even been dismissed as a placoderm synapomorphy by proponents of placoderm monophyly (Goujet, 2001), and is therefore absent PROBLEMS WITH PLACODERM PARAPHYLY from any of the recent lists of placoderm synapo- Although we currently favour a paraphyletic arrange- morphies (e.g. Goujet & Young, 2004; Young, 2008, ment for placoderms, we do not contend that this is 2010). More important is that these characters without problems. Recent proposals of placoderm are incongruent with a number of other features of paraphyly have highly pectinate arrangements, and

© 2014 The Authors. Zoological Journal of the Linnean Society published by John Wiley & Sons Ltd on behalf of The Linnean Society of London, Zoological Journal of the Linnean Society, 2014, 170, 779–821 EARLY GNATHOSTOME CHARACTERS 809 each disagrees in some significant details (Brazeau, placoderms are separated from osteichthyans by a 2009; Davis et al., 2012; Zhu et al., 2013). Common paraphyletic array of micromeric/tessellate taxa, then to all three hypotheses is that ptyctodonts and the support for this placement of arthrodires is at arthrodires do not form a clade, but there is disagree- least partially artefactual. ment over which is closest to the gnathostome crown. Resolving the question of placoderm monophyly Exceptional evidence of both claspers and internal will not come about through list-building projects fertilization has recently been discovered in these two with the objective of showing that Placodermi is a groups (Long et al., 2008; Ahlberg et al., 2009; Long, clade. As the only mandibulate stem gnathostomes, Trinajstic & Johanson, 2009). The phylogenetic placoderms are morphologically peculiar and care hypotheses of Brazeau (2009), Davis et al. (2012), and must be taken not to confuse their peculiarities Zhu et al. (2013) would therefore suggest that either with synapomorphies. Furthermore, dismissing com- this trait was acquired multiple times within the parisons between placoderms and osteostracans, gnathostomes or that it was lost in the osteichthyan galeaspids, and pituriaspids is counterproductive. We stem (see Ahlberg, 2009). It seems unparsimonious to see a more rigorous and careful comparative anatomy imply three acquisitions of this trait when two are of placoderms and agnathans to be a potential avenue already implied (chondrichthyans and ptyctodonts). for resolving these problems. Furthermore, there is no precedent in vertebrates for a transition from internal fertilization to external fertilization (see Long et al., 2009). However, on present evidence, it may simply suffice to solve this THE ‘ACANTHODIAN PROBLEM’ AND CHARACTERS OF dilemma by making arthrodires and ptyctodonts THE CHONDRICHTHYAN TOTAL GROUP a clade – but this solution was not recovered by As with Placodermi, Acanthodii has been unsup- Brazeau (2009), Davis et al. (2012), or Zhu et al. ported in nearly all published cladistic tests of its (2013). Furthermore, it conflicts with the peta- monophyly. However, there is some disagreement lichthyid like cranial morphology of ptyctodonts. about the precise stem placement of all of the Semidentine can be interpreted as intermediate acanthodian genera included in published clado- in type between mesodentine and orthodentine. grams. Davis et al. (2012) and Brazeau (2009) placed Semidentine shares polarization of the odontoblasts Acanthodes and its nearest relations on the within the mineralized matrix. However, although osteichthyan stem. However, the relationships of the orthodentine is found in chondrichthyans and osteich- Climatius-like acanthodians (sensu Brazeau, 2012) thyans (Sire et al., 2009), there may be reason on remain in conflict. Brazeau (2009) recovered most of the basis of character distributions to see these these taxa as stem chondrichthyans, whereas Davis as potentially independently derived, especially et al. (2012) resolved them as stem gnathostomes. if acanthodians are not monophyletic. Placoderm This lability of placement has two likely causes. One paraphyly not only requires the independent acquisi- is the small number of chondrichthyan total-group tion of odontoblast polarization, but also the rever- synapomorphies, the other is the small number of sion to mesodentine-type tissue. It is for this reason morphological characters that can be reliably scored that semidentine has been preserved, albeit with for acanthodians owing to the incompleteness of their reservations, in our list of potential placoderm fossils. synapomorphies. Remarkably few characters consistently identify A further concern relates to the dermal bones of members of the chondrichthyan total group under placoderms and osteichthyans and existing parsi- alternative placements of acanthodians. Of the char- mony software. The matrices of Brazeau (2009) acters that we have listed above, tessellate prismatic and Davis et al. (2012) employ a contingent coding calcified cartilage and the precerebral fontanelle (or reductive coding) technique (Hawkins, 2000). are fairly uncontroversial. Neither has ever been Although this limits the number of redundant states observed in an acanthodian. The phylogenetically and spurious effects of other coding methods, it is late appearance of individualized tooth bases rela- prone to length overestimates because of the limita- tive to a robustly calcified endoskeleton and the tions of currently available parsimony software precerebral fontanelle, as based on conditions in (Maddison, 1993). Even though the topologies aris- Doliodus, makes this character unlikely to help ing from analyses of these two data sets imply identify less crownward stem chondrichthyans. This nonhomology of placoderm and osteichthyan dermal leaves areally growing scales as the only feature cranial roofs, characters concerning this system unambiguously uniting any acanthodian-like fossils (such as the rectilinear skull roof pattern of (e.g. Kathemacanthus and Seretolepis from the arthrodires and osteichthyans) may be attracting MOTH locality, Hanke & Wilson, 2010) with the arthrodires towards the gnathostome crown. If chondrichthyan stem.

What phylogenetic conclusions can be drawn from the acquisition of the anal fin spines and no losses, this, and what are their implications in light of our Fig. 12E). In this hypothesis, Kathemacanthus and synapomorphy scheme? Reports of areally growing Seretolepis are chondrichthyans, but no more crown- scales in Climatius and Parexus (Burrow & Turner, ward than any other acanthodian. 2010; Burrow et al., 2013) have not been matched This latter phylogenetic hypothesis could be with stem-chondrichthyan identifications that have rejected if, for example, the full spine complement of been offered for both fragmentary and articulated Doliodus or Pucapampella included anal fin spines. taxa showing the same characters (Hanke & Wilson, Nevertheless, if only the scale growth type and anal 2010). Instead, they have been united with other fin spine characters are considered, then the most acanthodians on the basis of the scapulocoracoid parsimonious solution involving Kathemacanthus and characters that we consider spurious evidence of Seretolepis as stem chondrichthyans also places them acanthodian monophyly. However, let us give benefit in a clade with all known acanthodians. However, of the doubt to acanthodian monophyly and the we see no reason to separate some acanthodian-like identification of Kathemacanthus and articulated taxa as ‘putative chondrichthyans’ to the exclusion of Seretolepis as chondrichthyans (Fig. 12). We might others on the basis of such a small character sample then assume that acanthodians are either stem and without consideration of the diversity of outgroup osteichthyans or less crownward stem chondrich- conditions. thyans (this is never explicitly stated by acanthodian Acanthodian anatomy, and to a lesser extent early workers, but we must assume that it is one of these). chondrichthyan anatomy, is very poorly known. What The anal fin spine (the potential synapomorphy is known of acanthodians is mostly restricted to the that we retain above) could not be considered an dermal hard parts with few comparators in other acanthodian synapomorphy, but has two possible gnathostome taxa. However, approximately half of resolutions: either a gnathostome or chondrichthyan the characters in our synapomorphy scheme concern symplesiomorphy in these respective circumstances, some aspect of the endoskeleton, especially the or independently derived (Fig. 12A). The choice neurocranium. What is known of acanthodian depends on whether taxa such as Kathemacanthus neurocrania reveals widely divergent morpholo- and Seretolepis are paraphyletic or monophyletic gies, consistent with a nonmonophyletic Acanthodii (Fig. 12 shows optimal situations). However, similar (Brazeau, 2009). Phylogenetically significant axial interpretations must then follow the pattern of areal skeleton characters are known only for a small number scale growth (Fig. 12A–C). of acanthodids (e.g. Miles, 1970). This dearth of Choosing between these hypotheses depends on the endoskeletal data leaves few traits that can be com- placement of a presumed monophyletic Acanthodii. pared generally throughout the gnathostomes. If they are stem osteichthyans, then areally grow- The acanthodian problem may be resolved in very ing scales represent a generalized feature of the limiting and incremental steps as palaeontologists gnathostome crown (Fig. 12A–C), later modified mine the diversity and distribution of their hard higher on the osteichthyan stem and in acanthodians tissue anatomy. This is a necessary step towards a (where concentric superposition is acquired). The stable systematics of early gnathostomes. However, character therefore loses force as a chondrichthyan by itself it will probably be insufficient. As new synapomorphy, and the placement of Kathemacanthus acanthodian fossils rarely reveal anatomical details of and Seretolepis is Gnathostomata incertae sedis the endoskeleton, the rate of new taxonomic discov- (Fig. 12C). eries potentially outstrips the rate at which new If all acanthodians are stem chondrichthyans, this characters with links to non-acanthodians are discov- might imply that areally growing scales are actually ered. This means that it will become increasingly a plesiomorphic trait of acanthodians, later substi- difficult to add significantly more phylogenetically tuted by the concentric appositional growth more informative characters, which are needed to resolve ‘typical’ of acanthodian scales (Fig. 12D, E). All data sets with large amounts of missing data (Wiens, that ties Seretolepis and Kathemacanthus to the 2003a, b). Resolving the acanthodian problem will chondrichthyan stem in the hypothesis of Hanke & therefore require the discovery of uniquely well- Wilson (2010) is simply this character (‘Seretolepis- preserved acanthodian and other early gnathostome type’ areal scale growth). However, we are then endoskeletons. forced to propose the acquisition and subsequent loss of the anal fin and admedian fin spines along CONCLUSIONS the chondrichthyan stem. This makes placing Kathemacanthus and Seretolepis as the immediate We have created a nested hierarchical synapomor- sister groups of all acanthodians the parsimonious phy scheme for the deepest crown nodes of the solution for both characters (leaving us with only Gnathostomata using an outgroup-based approach

AB Parexus Parexus + +

Seretolepis Seretolepis CHONDRICHTHYESKathemacanthus+ ACANTHODII Climatius OSTEICHTHYES CHONDRICHTHYESKathemacanthus+ ACANTHODII Climatius OSTEICHTHYES Anal fin Areally growing spine Areally growing Anal fin spine (A) Areally growing scales scales scales Areally growing Anal fin spine Anal fin spine scales (D) 5 steps 5 steps Anal fin spine (A)

Areally growing scales Areally growing scales

CD Parexus Parexus + +

Seretolepis Seretolepis CHONDRICHTHYESKathemacanthus+ ACANTHODII Climatius OSTEICHTHYES CHONDRICHTHYESKathemacanthus+ ACANTHODII Climatius OSTEICHTHYES Areally growing Anal fin spine (A) Areally growing scales Areally growing scales Anal fin spine scales (D) 4 steps Areally growing scales Anal fin spine (A) Areally growing scales

EF Parexus Parexus + + Seretolepis Seretolepis Kathemacanthus+ Kathemacanthus+ Climatius CHONDRICHTHYES Climatius ACANTHODII OSTEICHTHYES CHONDRICHTHYES ACANTHODII OSTEICHTHYES

Areally growing Anal fin spine* Anal fin spine Anal fin spine* scales Anal fin spine Anal fin spine Areally growing scales Areally growing scales 3 steps Presence 3 or 4* steps Absence / Ambiguous gain/loss Anal fin spine* (A) Accelerated transformation (D) Delayed transformation

Figure 12. Alternative phylogenetic placements for problematic acanthodian-like taxa and their implications in light of two characters discussed in the text and assumptions of acanthodian monophyly. A, one of the hypotheses implied by placing only Kathemacanthus and Seretolepis on the chondrichthyan stem, independent of other acanthodians. B, equally parsimonious placement of Kathemacanthus and Seretolepis on osteichthyan stem. C, consensus tree showing that resolution of Kathemacanthus and Seretolepis to the chondrichthyan stem collapses if all other acanthodians are placed on the osteichthyan stem (areal scale growth is plesiomorphic). D–F, improvements to parsimony score if some or all acanthodians are moved to the chondrichthyan stem (restoration of areal scale growth as a chondrichthyan synapomorphy). D, hypothesis in which all acanthodians are stem chondrichthyans, but paraphyletic. E, hypothesis in which all acanthodians are stem chondrichthyans but monophyletic. F, hypothesis in which taxa with areally growing scales are stem chondrichthyans, whereas the remaining (assumed monophyletic) Acanthodii are stem osteichthyans. Ambiguities in character state distributions based on a soft polytomy may entail different lengths depending on their resolution. Note that loss of areal scale growth in Acanthodii and Osteichthyes reﬂects transitions to different, not identical, states and must therefore be treated as separate events. Asterisk indicates values derived from resolving stem chondrichthyan polytomy as a paraphylum with respect to the crown.

(Fig. 13). Our preferred solution places placoderms 2009; Davis et al., 2012; Zhu et al., 2013). This hinges as a paraphyletic array of stem gnathostomes. on the absence of pelvic ﬁns, which is likely to We have omitted antiarchs, even though most reﬂect a derived state in the group (Zhu et al., 2012b). phylogenetic analyses place them as the sister group They might therefore act as a ‘wildcard’ taxon of all other mandibulate gnathostomes (Brazeau, within the gnathostome stem. We suggest two

Seretoleps +

Parexus +

OSTEOSTRACIBrindabellaspisMacropetalichthysARTHRODIRAEntelognathus CHONDRICHTHYESKathemacanthusClimatiusISCHNACANTHIFORMESACANTHODIFORMESOSTEICHTHYES

32, 35, 37 49-51, 53-56, 57-64

34 (29, 30, 36) 40-44, 47, 48, 57 (45, 46, 52)

A 19-26, 31, 33, 38, 39 10, 11, 12, (27)

14 (7, 8, 9, 13, 15, 16)

1-6

Seretoleps +

Parexus +

OSTEOSTRACIBrindabellaspisMacropetalichthysARTHRODIRAEntelognathus CHONDRICHTHYESKathemacanthusClimatiusISCHNACANTHIFORMESACANTHODIFORMESOSTEICHTHYES

32, 35, 37

31, 33, 34, 38, 39 (29, 30, 36) 41, 45, 46, 49, 51-56, 59, 62-64

B 19-26 10, 11, 12, (27)

14 (7, 8, 9, 13, 15, 16)

1-6

Figure 13. Summary cladograms of hypotheses of phylogenetic placements argued in this paper. A, cladogram depicting acanthodian genera distributed on the chondrichthyan and osteichthyan stems. B, cladogram depicting acanthodians restricted to chondrichthyan stem, but left unresolved. Character transformation labels at internal nodes correspond to those in the text. Numbers in parentheses reﬂect ambiguities that are resolved to their most inclusive level (i.e. ‘accelerated transformation’) and could have more restricted distributions. alternative hypotheses concerning the placement endoskeletal characters, as well as some braincase of acanthodians. The ﬁrst of these places most characters in Acanthodes newly recognized by Davis acanthodians on the chondrichthyan stem, with et al. (2012) and evidence for the homology of Acanthodes and its closest relatives representing placoderm and osteichthyan macromeric skull condi- members of the osteichthyan total group (Fig. 13A). tions described by Zhu et al. (2013). These alterna- The second hypothesis places all acanthodians as tives differ from one another by only two steps under stem chondrichthyans (Fig. 13B). This latter hypoth- our synapomorphy scheme, and we regard this as esis enjoys support from a number of dermal and insufficient to distinguish between them at present.

We have emphasized the importance of phyloge- Anderson PSL. 2008. Cranial muscle homology across netic background assumptions in argumentation modern gnathostomes. Biological Journal of the Linnean about the systematic significance of characters under- Society 94: 195–216. writing this set of tree topologies. This explicitness Anderson PSL, Friedman M, Brazeau MD, Rayfield EJ. renders our proposals open to empirical refutation in 2011. Initial radiation of jaws demonstrated stability the hope that this work will promote future research despite faunal and environmental change. Nature 476: 206– focused on testing these distributions in a methodo- 209. logically consistent manner. We expect that many of Andrews SM, Long JA, Ahlberg PE, Barwick R, Campbell K. 2005. The structure of the sarcopterygian Onychodus these proposals will succumb to refutation, consistent jandemarrai n. sp. from Gogo, Western Australia: with a with the current level of uncertainty in the study of functional interpretation of the skeleton. Transactions of the early gnathostome phylogenetics. Because of this, we Royal Society of Edinburgh 96: 197–307. have not only emphasized our current preferred solu- Bernacsek GM, Dineley DL. 1977. New acanthodians tion, but have explored a number of alternatives that from the Delorme Formation (Lower Devonian) of N.W.T., could easily supplant our proposal with further study Canada. Palaeontographica A 159: 1–25. and new fossil data. Beverdam A, Merlo GR, Paleari L, Mantero S, Genova F, Barbieri O, Janvier P, Levi G. 2002. Jaw transformation ACKNOWLEDGEMENTS with gain of symmetry after Dlx5/Dlx6 inactivation: mirror of the past? Genesis 34: 221–227. We thank Zerina Johanson (NHM, London), Stig Blair JE, Hedges SB. 2005. Molecular phylogeny of diver- Walsh (NMS, Edinburgh), and Mark Wilson (UALVP) gence times of deuterostome animals. Molecular Biology for access to collections. Comments from two anony- and Evolution 22: 2275–2284. mous reviewers helped improve an earlier draft and Blieck A, Heintz N. 1983. The cyathaspids of the Red Bay Zerina Johanson provided helpful comments on the Group (Lower Devonian) of Spitsbergen. Polar Research 1: revised text. For reviews of the penultimate draft, we 49–74. thank two anonymous referees and, especially, Rob Blom H, Märss T. 2010. The interrelationships and evolu- Sansom who pointed out critical errors in some of the tionary history of anaspids. In: Elliott DK, Maisey JG, character optimizations in figures. M. D. B. thanks Yu X, Miao D, eds. Morphology, phylogeny and paleobiogeography of fossil fishes. Munich: Verlag Dr. Friedrich Pfeil, support from European Research Council under the 45–58. European Union’s Seventh Framework Programme Bock WJ. 1969. Discussion: the concept of homology. Annals (FP/2007–2013)/ERC Grant Agreement number of the New York Academy of Sciences 167: 71–73. 311092; and support from Naturalis Biodiversity Brazeau MD. 2009. The braincase and jaws of a Devonian Center. M. F. was supported by the John Fell Fund of ‘acanthodian’ and modern gnathostome origins. Nature 457: the University of Oxford and St Hugh’s College. 305–308. Brazeau MD. 2012. A revision of the anatomy of the Early REFERENCES Devonian jawed vertebrate Ptomacanthus anglicus Miles. Palaeontology 55: 355–367. Adrain JM, Wilson MVH. 1994. Early Devonian Burrow CJ. 2004. Acanthodian fishes with dentigerous jaw cephalaspids (Vertebrata: Osteostraci: Cornuata) from the bones: the Ischnacanthiformes and Acanthodopsis. Fossils southern Mackenzie mountains, N. W. T., Canada. Journal and Strata 50: 8–22. of Vertebrate Paleontology 14: 301–319. Burrow CJ, Jones AS, Young GC. 2005. X-ray Agassiz L. 1833–1844. Recherches sur les Poissions Fossiles. micotomography of 410 million-year-old optic capsules from V vols. Neuchâtel: Imprimerie de Petitpierre et Prince. placoderm fishes. Micron 36: 551–557. Agassiz L. 1844–1845. Monographie des Poissons Fossiles du Burrow CJ, Newman MJ, Davidson RG, Den Blaauwen Vieux Grès Rouge ou Système Dévonien (Old Red Sandstone) JL. 2011. Sclerotic plates or circumorbital bones in early des Iles Britanniques et de Russie. Neuchâtel: Soleure, chez jawed fishes? Palaeontology 54: 207–214. Jent & Gassman. Burrow CJ, Newman MJ, Davidson RG, Den Blaauwen Ahlberg PE. 2009. Palaeontology: birth of jawed vertebrates. JL. 2013. Redescription of Parexus recurvus, an Early Devo- Nature 457: 1094–1095. nian acanthodian from the Midland Valley of Scotland. Ahlberg PE, Trinajstic K, Johanson Z, Long JA. 2009. Alcheringa 37: 1–23. Pelvic claspers confirm chondrichthyan-like internal fertili- Burrow CJ, Trinajstic K, Long JA. 2012. First acanthodian zation in arthrodires. Nature 460: 888–889. from the Upper Devonian (Frasnian) Gogo Formation, Allis EP. 1897. The cranial muscles, and cranial and first Western Australia. Historical Biology 4: 349–357. spinal nerves in Amia calva. Journal of Morphology 12: Burrow CJ, Turner S. 1999. A review of placoderm scals, 487–808. and their significance in placoderm phylogeny. Journal of Allis EP. 1923. The cranial anatomy of Chlamydoselachus Vertebrate Paleontology 19: 204–219. anguineus. Acta Zoologica 4: 123–221. Burrow CJ, Turner S. 2010. Reassessment of ‘Protodus’

scoticus from the Early Devonian of Scotland. In: Elliott DK, De Beer GR, Moy-Thomas JA. 1935. On the skull of Maisey JG, Yu X, Miao D, eds. Morphology, phylogeny and Holocephali. Philosophical Transactions of the Royal Society paleobiogeography of fossil fishes. Munich: Verlag Dr. of London, Series B 224: 287–312. Friedrich Pfeil, 123–144. De Pinna MCC. 1991. Concepts and tests of homology in the Burrow CJ, Turner S. 2013. Scale structure of the puta- cladistic paradigm. Cladistics 7: 367–394. tive chondrichthyan Gladbachus adentatus Heidtke & Dean B. 1895. Fishes living and fossil. An outline of their Krätschmer, 2001 from the Middle Devonian Rheinisches forms and possible relationships. New York: MacMillan and Schiefergebirge, Germany. Historical Biology 25: 385–390. Co. Caldwell MC, Wilson MVH. 1995. Comparison of the body Dean B. 1899. The so-called Devonian ‘Lamprey’, Palaeospon- form and squamation of ‘fork-tailed’ agnathans with that of dylus gunni Trq., with notes on the systematic arrangement conventional thelodonts. Geobios M.S. 19: 23–29. of the fish-like vertebrates. Memoirs of the New York Carr R, Lelièvre H, Jackson GL. 2010. The ancestral Academy of Sciences 2: 1–32. morphotype for the gnathostome pectoral fin revisited. Dean B. 1901. Palaeontological notes. I. On two new In: Elliott DK, Maisey JG, Yu X, Miao D, eds. Morphology, arthrodires from the Cleveland Shale of Ohio. II. On the phylogeny and paleobiogeography of fossil fishes. Munich: characters of Mylostoma Newberry. III. Further notes on the Verlag Dr. Friedrich Pfeil, 107–122. relationships of the Arthrognathi. Memoirs of the New York Chen M, Zhou M, Yang L, He S. 2012. Basal jawed verte- Academy of Sciences 2: 87–124. brate phylogenomics using transcriptomic data from Solexa Dean B. 1907. Notes on acanthodian sharks. American sequencing. PLoS ONE 7: e36256. Journal of Anatomy 7: 209–226. Clark AJ, Summers AP. 2007. Morphology and kinematics Dean B. 1909. Studies on fossil fishes (sharks, chimaeroids of feeding in hagfish: possible functional advantages of jaws. and arthrodires). Memoirs of the American Museum of Journal of Experimental Biology 210: 3897–3909. Natural History 9: 209–287. Clement JG. 1992. Re-examination of the fine structure Dean MN, Summers AP. 2006. Mineralized cartilage in the of endoskeletal mineralization in Chondrichthyans: skeleton of chondrichthyans fishes. Zoology 109: 164–168. implications for growth, ageing and calcium homeostasis. Denison RH. 1967. Ordovician vertebrates from western Australian Journal of Marine and Freshwater Research 43: United States. Fieldiana, Geology 16: 131–192. 157–181. Denison RH. 1975. Evolution and classification of placoderm Coates MI, Gess R. 2007. A new reconstruction of fishes. Breviora 432: 1–24. Onychoselache traquari, comments on early chondrichthyan Denison RH. 1978. Placodermi. In: Schultze H-P, ed. Hand- pectoral girdles and hybodontiform phylogeny. Palaeontol- book of paleoichthyology. Vol. 2. Stuttgart: Gustav Fischer ogy 50: 1421–1446. Verlag, 1–128. Coates MI, Sequeira SEK. 1998. The braincase of a primi- Denison RH. 1979. Acanthodii. In: Schultze H-P, ed. Hand- tive shark. Transactions of the Royal Society of Edinburgh: book of paleoichthyology. Vol. 5. Stuttgart: Gustav Fischer Earth Sciences 89: 63–85. Verlag, 1–62. Coates MI, Sequeira SEK. 2001. A new stethacanthid Dick JRF. 1981. Diplodoselache woodi gen. et sp. nov., chondrichthyans from the Lower Carbonifeous of Bearsden, an early Carboniferous shark from the Midland Valley of Scotland. Journal of Vertebrate Paleontology 21: 438–459. Scotland. Transactions of the Royal Society of Edinburgh: Coates MI, Sequeira SEK, Sansom IJ, Smith MM. 1998. Earth Sciences 72: 99–113. Spines and tissues of ancient sharks. Nature 396: 729–730. Dineley DL, Loeffler EJ. 1976. Ostracoderm faunas of the Cope ED. 1889. Synopsis of the families of Vertebrata. Ameri- Delorme and associated Siluro-Devonian formations, North can Naturalist 23: 849–877. West Territories, Canada. Special Papers in Palaeontology Cope ED. 1892. On the phylogeny of Vertebrata. Proceedings 18: 1–214. of the American Philosophical Society 30: 278–281. Donoghue PCJ, Forey PL, Aldridge RJ. 2000. Conodont Crane A. 1877. On certain living genera of fishes and their affinity and chordate phylogeny. Biological Reviews 75: 191– fossil affinities. Geological Magazine 4: 209–219. 251. Cunningham JA, Rücklin M, Blom H, Botella H, Donoghue PCJ, Purnell MA. 1999. Mammal-like occlusion Donoghue PCJ. 2012. Testing models of development in in conodonts. Paleobiology 25: 58–74. the earliest bony vertebrates, Andreolepis and Lophosteus. Donoghue PCJ, Sansom IJ, Downs JP. 2006. Early evo- Biology Letters 8: 833–837. lution of vertebrate skeletal tissues and cellular interac- Davis JW. 1894. On the fossil fish-remains of the Coal tions, and the canalization of skeletal development. Journal Measures of the British Islands, 1: Pleuracanthidae. Scien- of Experimental Zoology 306B: 278–294. tific Transactions of the Royal Dublin Society 2: 703–748. Donoghue PCJ, Smith PM. 2001. The anatomy of Davis SP, Finarelli JA, Coates MI. 2012. Acanthodes and Turinia pagei (Powrie), and the phylogenetic status of the shark-like conditions in the last common ancestor of modern Thelodonti. Transactions of the Royal Society of Edinburgh: gnathostomes. Nature 486: 247–250. Earth Sciences 92: 15–37. De Beer GR. 1931. The development of the skull of Scyllium Elliott DK, Carr RK. 2010. A new species of Bryantolepis (Scyliorhinus) canicula L. Quarterly Journal of Microscopi- Camp, Welles, and Green, 1949 (Placodermi, Arthrodira) cal Science new series 74: 591–646. from the Early Devonian Water Canyon Formation of

northern Utah and southern Idaho, with comments on the Gill T. 1872. Arrangement of the families of fishes, or classes endocranium. Kirtlandia 57: 22–35. Pisces, Marsipobranchii, and Leptocardii. Smithsonian Mis- Farris JS. 1982. Outgroups and parsimony. Systematic cellaneous Collections 247: 1–49. Zoology 31: 328–334. Gillis JA, Rawlinson KA, Bell J, Lyon WS, Barker CVH, Forey PL, Gardiner BG. 1986. Observations on Ctenurella Shubin NH. 2011. Holocephalan embryos provide evidence (Ptyctodontida) and the classification of placoderm fishes. for gill arch appendage reduction and opercular evolution in Zoological Journal of the Linnean Society 86: 43–74. cartilaginous fishes. Proceedings of the National Academy of Forey PL, Janvier P. 1993. Agnathans and the origin of Sciences 108: 1507–1512. jawed vertebrates. Nature 361: 129–134. Goodrich ES. 1909. Vertebrata Craniata (first fascicle: Franz-Odendaal TA. 2011. The ocular skeleton through the cyclostomes and fishes. In: Lankester R, ed. Treatise on eye of evo-devo. Journal of Experimental Biology 316B: zoology. Part 9. London: Adam and Charles Black. 393–401. Goujet D. 1973. Sigaspis, un novel Arthrodire du Dévonien Friedman M. 2007a. Styloichthys as the oldest coelacanth: Inférieur du Spitsberg. Palaeontographica 143A: 73–88. implications for early osteichthyan interrelationships. Goujet D. 1982. Les affinités des placodermes, une revue Journal of Systematic Palaeontology 5: 289–343. des hypothèses actuelles. Geobios Mémoire Spécial 6: Friedman M. 2007b. The interrelationships of Devonian 27–38. lungfishes (Sarcopterygii: Dipnoi) as inferred from Goujet D. 1984a. Les Poissons placodermes du Spitsberg. neurocranial evidence and new data from the genus Arthrodires Dolichothoraci de la Formation de Wood Bay Soederberghia Lehman, 1959. Zoological Journal of the (Dévonien inférieur). Cahiers de Paléontologie (section Linnean Society 151: 115–171. Vertébrés). Paris: Éditions du CNRS. Friedman M, Brazeau MD. 2010. A reappraisal of the origin Goujet D. 1984b. Placoderm interrelationships: a new inter- and basal radiation of the Osteichthyes. Journal of Verte- pretation, with a short review of placoderm classifications. brate Paleontology 30: 36–56. Proceedings of the Linnean Society of New South Wales 107: Friedman M, Brazeau MD. 2013. Palaeontology: a jaw- 211–243. dropping fossil fish. Nature 502: 175–177. Goujet D. 2001. Placoderms and basal gnathostome Fritsch AJ. 1890. Fauna der Gaskohle und der Kalksteine apomorphies. In: Ahlberg PE, ed. Major events in early der Permformation Böhmens. Bd. 3, Heft 1, Selachii vertebrate evolution: palaeontology, phylogeny, genetics and (Pleuracanthus, Xenacanthus). Prague: Selbstverlag. development. London: Taylor & Francis, 209–222. Gagnier P-Y. 1993. Sacabambaspis janvieri, Vertébré Goujet D, Janvier P, Suarez-Riglos M. 1985. Un nouveau ordovicien de Bolivie. 1. Analyse morphologique. Annales de rhénanide (Vertebrata, Placodermi) de la Formation Belén Paléontologie 79: 19–69. (Dévonien moyen), Bolivie. Annales de Paléontologie 71: Gagnier P-Y, Hanke GF, Wilson MVH. 1999. Tetanopsyrus 35–53. lindoei, gen. et sp. nov., an Early Devonian acanthodian Goujet D, Young GC. 1995. Interrelationships of placoderms from the Northwest Territories, Canada. Acta Geologica revisited. Geobios Mémoire Spécial 19: 89–95. Polonica 49: 81–96. Goujet D, Young GC. 2004. Placoderm anatomy and Gagnier P-Y, Paris F, Racheboeuf P, Janvier P, phylogeny: new insights. In: Arratia G, Wilson MVH, Suarez-Riglos M. 1989. Les vertebres de Bolivie: donnés Cloutier R, eds. Recent advances in the origin and early biostratigraphiques et anatomiques complémentaires. Bul- radiation of vertebrates. Munich: Verlag Dr Friedrich Pfeil, letin de l’Institut Français d’Études Andines 18: 75–93. 109–126. Gagnier P-Y, Wilson MVH. 1996. Early Devonian Gross W. 1931. Asterolepis ornata Eichw. und das Antiarchi- acanthodians from northern Canada. Paleontology 39: 241– Problem. Palaeontographica 75A: 1–62. 258. Gross W. 1937. Die Wirbeltiere des rheinischen Devons. Gai Z, Donoghue PCJ, Zhu M, Janvier P, Stampanoni M. Abhandlungen der Preussische Geologische Landesanstalt, 2011. Fossil jawless fish from China foreshadows early Neue Folge 76: 1–83. jawed vertebrate anatomy. Nature 476: 324–327. Gross W. 1961. Lunaspis broilii und Lunaspis heroldi aus dem Gardiner BG. 1984a. The relationship of placoderms. Hunsrück-schiefer (Unterdevon, Rheinland). Notizblatt des Journal of Vertebrate Paleontology 4: 379–395. Hessischen Landesamtes für Bodenforschung 89: 17–43. Gardiner BG. 1984b. The relationships of the palaeoniscid Gross W. 1962. Neuuntersuchung der Stensiöellida fishes, a review based on new specimens of Mimia and (Arthrodira, Unterdevon). Notizblatt des Hessischen Moythomasia from the Upper Devonian of Western Aus- Landesamtes für Bodenforschung 89: 17–43. tralia. Bulletin of the British Museum (Natural History): Gross W. 1968. Porenschuppen und Sinneslinien des Geology 37: 173–428. Thelodontiers Phlebolepis elegans Pander. Paläontologische Gess RW, Coates MI, Rubidge BS. 2006. A lamprey from Zeitschrift 42: 131–146. the Devonian period of South Africa. Nature 443: 981–984. Gudo M, Homberger DG. 2002. The functional morphology Giles S, Rücklin M, Donoghue PCJ. 2013. Histology of of the pectoral fin girdle of the spiny dogfish (Squalus ‘placoderm’ dermal skeletons: implications for the nature of acanthias): implications for the evolutionary history of the the ancestral gnathostome. Journal of Morphology 274: pectoral girdle of vertebrates. Senckenbergiana Lethaea 82: 627–644. 241–252.

Halstead LB. 1973. The heterostracan fishes. Biological Heintz A. 1967. A new tremataspidid from Ringerike, south Reviews 48: 279–332. Norway. Journal of the Linnean Society of London (Zoology) Halstead LB. 1979. Internal anatomy of the polybrachaspids 47: 55–68. (Agnatha, Galeaspida). Nature 282: 833–836. Hemmings SK. 1978. The Old Red Sandstone antiarchs of Hamel M-H, Poplin C. 2008. The braincase anatomy of Scotland: Pterichthyodes and Microbrachius. Palaeonto- Lawrenciella schaefferi, actinopterygian from the Upper graphical Society (Monograph) 131: 1–64. Carboniferous of Kansas (USA). Journal of Vertebrate Heyler D. 1958. Remarques sur la caudale d’Acanthodes. Paleontology 28: 989–1006. Compte rendu hebdomadaire des seances de l’Academie des Hanke GF. 2002. Paucicanthus vanelsti gen. et sp. nov., an Sciences Paris 247: 1636–1639. Early Devonian (Lochkovian) acanthodian that lacks paired Heyler D. 1962. Les acanthodiens et le problème de fin-spines. Canadian Journal of Earth Sciences 39: 1071– l’Aphetohyoïde. Colloques internationaux du Centre national 1083. de la Recherche scientifique 104: 39–47. Hanke GF. 2008. Promesacanthus eppleri n. gen. et n. sp., a Heyler D. 1969. Acanthodii. In: Piveteau J, ed. Traité de mesacanthid (Acanthodii, Acanthodiformes) from the Lower Paléontologie. 4 Paris: Masson, 21–70. Devonian of northern Canada. Geodiversitas 30: 287–302. Holmgren N. 1942. Studies on the head of fishes. Part III. Hanke GF, Davis SP. 2008. Redescription of the Acta Zoologica 23: 129–261. acanthodian Gladiobranchus probaton Bernacsek & Hussakof L. 1905. Notes on the Devonian ‘Placoderm’, Dineley, 1977, and comments on diplacanthid relationships. Dinichthys intermedius Nwb. Bulletin of the American Geodiversitas 30: 303–330. Museum of Natural History 21: 27–36. Hanke GF, Davis SP. 2012. A re-examination of Lupopsyrus Hussakof L. 1906. Studies on the Arthrodira. Memoires of the pygmaeus Bernacsek & Dineley, 1977 (Pisces, Acanthodii). American Museum of Natural History 3: 103–154. Geodiversitas 34: 469–487. Huxley TH. 1861. Preliminary essay upon the systematic Hanke GF, Davis SP, Wilson MVH. 2001. New species arrangement of the fishes of the Devonian epoch. Memoirs – of the acanthodian genus Tetanopsyrus from northern Geological Survey of the United Kingdom 10: 1–40. Canada, and comments on related taxa. Journal of Verte- Janvier P. 1975. Les yeux des Cyclostomes fossiles et le brate Paleontology 21: 740–753. problème de l’origine des Myxinoïdes. Acta Zoologica 56: Hanke GF, Wilson MVH. 2004. New teleostome fishes and 1–9. acanthodian systematics. In: Arratia G, Wilson MVH, Janvier P. 1981a. The phylogeny of the Craniata, with Cloutier R, eds. Recent advances in the origin and early particular reference to the significance of fossil ‘agnathans’. radiation of vertebrates. Munich: Verlag Dr Friedrich Pfeil, Journal of Vertebrate Paleontology 1: 121–159. 189–216. Janvier P. 1981b. Norselaspis glacialis n.g., n.sp. et les Hanke GF, Wilson MVH. 2006. Anatomy of the Early Devo- relations phylogénétiques entre les Kiaeraspidiens (Osteo- nian acanthodian Brochoadmones milesi based on nearly straci) du Dévonien inférieur du Spitsberg. Palaeovertebrata complete body fossils, with comments on the evolution 11: 19–131. and development of paired fins. Journal of Vertebrate Janvier P. 1984. The relationships of the Osteostraci Paleontology 26: 526–537. and Galeaspida. Journal of Vertebrate Paleontology 4: 344– Hanke GF, Wilson MVH. 2010. The putative stem-group 358. chondrichthyans Kathemacanthus and Seretolepis from the Janvier P. 1985a. Les céphalaspides du Spitsberg. Cahiers Lower Devonian MOTH locality, Mackenzie Mountains, de Paléontologie (section Vertébrés). Paris: Éditions du Canada. In: Elliott DK, Maisey JG, Yu X, Miao D, eds. CNRS. Morphology, phylogeny and paleobiogeography of fossil Janvier P. 1985b. Les thyestidiens (Osteostraci) du Siluren fishes. Munich: Verlag Dr. Friedrich Pfeil, 159–182. de Saaremaa (Estonie) première partie : morphologie et Hanke GF, Wilson MVH, Saurette FJ. 2013. Partial anatomie. Annales de Paléontologie 71: 83–147. articulated specimen of the Early Devonian putative Janvier P. 1993. Patterns in diversity in the skull of jawless chondrichthyan Polymerolepis whitei Karataju¯ te?-Talimaa, fishes. In: Hanke J, Hall BK, eds. The skull. Volume 2, 1968, with an anal fin spine. Geodiversitas 35: 529–543. patterns of structural and systematic diversity. Chicago, IL: Harris JE. 1951. Diademodus hydei, a new fossil shark from University of Chicago Press, 131–188. the Cleveland Shale. Proceedings of the Zoological Society of Janvier P. 1996a. Early vertebrates. Oxford: Clarendon London 120: 683–697. Press. Hawkins JA. 2000. A survey of primary homology assess- Janvier P. 1996b. The dawn of the vertebrates: characters ment: different botanists perceive and define characters versus common ascent in the rise of current vertebrate in different ways. In: Scotland RW, Pennington RT, eds. phylogenies. Palaeontology 39: 259–287. Homology and systematics: coding characters for Janvier P. 2001. Ostracoderms and the shaping of the phylogenetic analysis. London: Taylor & Francis, 22–53. gnathostome characters. In: Ahlberg PE, ed. Major events Heintz A. 1932. The structure of Dinichthys. A contribution to in early vertebrate evolution: palaeontology, phylogeny, genet- our knowledge of the Arthrodira. In: Gudger EW, ed. The ics and development. London: Taylor & Francis, 172–186. Bashford Dean memorial volume. Part 4. New York: Ameri- Janvier P. 2004. Early specializations in the branchial appa- can Museum of Natural History, 115–224. ratus of jawless vertebrates: a consideration of gill number

and size. In: Arratia G, Wilson MVH, Cloutier R, eds. Recent Liem KA, Summers A. 1999. Gross anatomy and functional advances in the origin and early radiation of vertebrates. morphology of muscles. In: Hamlett WC, ed. Sharks, skates Munich: Verlag Dr Friedrich Pfeil, 29–52. and rays: the biology of elasmobranch fish. Baltimore: Johns Janvier P, Arsenault M. 2002. Palaeobiology: calcification of Hopkins University Press, 93–114. early vertebrate cartilage. Nature 417: 609. Liu Y. 1975. Lower Devonian agnathans of Yunnan and Janvier P, Arsenault M, Desbiens S. 2004. Calcified car- Sichuan. Vertebrata PalAsiatica 10: 202–216. tilage in the paired fins of the osteostracan Escuminaspis Long JA. 1983. A new diplacanthoid acanthodian from the laticeps (Traquair, 1880), from the Late Devonian of Late Devonian of Victoria. Memoirs of the Association of Miguasha (Québec, Canada), with a consideration of the Australasian Palaeontologists 1: 51–65. early evolution of the pectoral fin endoskeleton in verte- Long JA. 1986. New ischnacanthid acanthodians from the brates. Journal of Vertebrate Paleontology 24: 773–779. Early Devonian of Australia, with comments on acanthodian Janvier P, Blieck A. 1979. New data on the internal interrelationships. Zoological Journal of the Linnean anatomy of the Heterostraci (Agnatha), with general Society 87: 321–339. remarks on the phylogeny of the Craniota. Zoologica Scripta Long JA. 1988. New palaeoniscoid fishes from the Late 8: 287–296. Devonian and Early Carboniferous of Victoria. Memoirs of Janvier P, Suarez-Riglos M. 1986. The Silurian and Devo- the Association of Australasian Palaeontologists 7: 1–64. nian vertebrates of Bolivia. Bulletin de l’Institut Français Long JA. 1997. Ptyctodontid fishes (Vertebrata, Placodermi) d’Études Andines 15: 74–114. from the Late Devonian Gogo Formation, Western Aus- Jarvik E. 1972. Middle and Upper Devonian Porolepiformes tralia, with a revision of the European genus Ctenurella from East Greenland with special reference to Glyptolepis Ørvig, 1960. Geodiversitas 19: 515–555. groenlandica n.sp. Meddelelser om Grønland 187: 1– Long JA, Barwick RE, Campbell KSW. 1997. Osteology 295. and functional morphology of the osteolepiform fish Jarvik E. 1977. The systematic position of acanthodian Gogonasus andrewsae Long, 1985, from the Upper Devonian fishes. In: Andrews SM, Miles RS, Walker AD, eds. Prob- Gogo Formation, Western Australia. Records of the Western lems in vertebrate evolution. London: Academic Press, 199– Australian Museum Supplement 53: 1–89. 225. Long JA, Trinajstic K, Johanson Z. 2009. Devonian Jarvik E. 1980. Basic structure and evolution of vertebrates. arthrodire embryos and the origin of internal fertilization in Volume 1. London: Academic Press. vertebrates. Nature 457: 1124–1127. Johanson ZA. 2002. Vascularization of osteostracan and Long JA, Trinajstic K, Young GC, Senden T. 2008. Live antiarch (Placodermi) pectoral fin: similarities, and impli- birth in the Devonian period. Nature 453: 650–652. cations for placoderm relationships. Lethaia 35: 169–186. Long JA, Young GC. 1988. Acanthothoracid remains from Johanson ZA, Smith MM. 2005. Origin and evolution of the Early Devonian of New South Wales, including a com- gnathostome dentitions: a question of teeth and pharyngeal plete sclerotic capsule and pelvic girdle. Memoirs of the denticles in placoderms. Biological Reviews 80: 303–345. Association of Australasian Palaeontologists 7: 65–80. Kade G. 1858. Ueber die devonischen Fischreste eines Lutken C. 1871. On the limits and classification of the ganoids. Diluvialblockes. Meseritz: F. W. Lorenz. Annals and Magazine of Natural History 7: 329–339. Karatajute-Talimaa V. 1998. Determination methods for the M’Coy F. 1848. On some new fossil fish of the Carboniferous exoskeletal remains of early vertebrates. Mitteilungen aus period. Annals and Magazine of Natural History 2: 1–10. dem Museum für Naturkunde in Berlin, Geowissenschaft- Maddison WP. 1993. Missing data versus missing characters liche Reihe 1: 21–52. in phylogenetic analysis. Systematic Biology 42: 576–581. Karatajute-Talimaa V, Smith MM. 2003. Early acantho- Maisey JG. 1986. Heads and tails: a chordate phylogeny. dians from the Lower Silurian of Asia. Transactions of the Cladistics 2: 201–256. Royal Society of Edinburgh: Earth Sciences 93: 277–299. Maisey JG. 1989. Hamiltonichthys mapesi g. & sp. nov. Keating JN, Sansom RS, Purnell MA. 2012. A new (Chondrichthyes; Elasmobranchii), from the Upper Pennsyl- osteostracan fauna from the Devonian of the Welsh border- vanian of Kansas. American Museum Novitates 2931: 1–42. lands and observations on the taxonomy and growth of Maisey JG. 2001. A primitive chondrichthyan braincase Osteostraci. Journal of Vertebrate Paleontology 32: 1002– from the Middle Devonian of Bolivia. In: Ahlberg PE, ed. 1017. Major events in early vertebrate evolution: palaeontology, Kemp NE, Westrin SK. 1979. Ultrastructure of calcified phylogeny, genetics and development. London: Taylor & cartilage in the endoskeletal tesserae of sharks. Journal of Francis, 263–288. Morphology 160: 75–102. Maisey JG. 2005. Braincase of the Upper Devonian shark Kner R. 1868. Ueber Conchopoma gadiforme und Acanthodes Cladodoides wildungensis (Chondrichthyes, Elasmobran- aus dem Rothliegended (der untern Dyas) von Lebach bei chii), with observations on the braincase in early Saarbrücken in Rheinprueussen. Sitzungsberichte der chondrichthyans. Bulletin of the American Museum of Akademie der Wissenschaften mathematisch-naturwissen- Natural History 288: 1–103. schaftliche Klasse 57: 278–304. Maisey JG. 2007. The braincase in Paleozoic symmoriiform Koentges G, Matsuoka T. 2002. Jaws of the fates. Science and cladoselachian sharks. Bulletin of the American 298: 371–373. Museum of Natural History 307: 1–122.

Maisey JG. 2012. What is an ‘elasmobranch’? The impact of Moy-Thomas JA. 1936. On the structure and affinities of palaeontology in understanding elasmobranch phylogeny the Carboniferous cochliodont Helodus simplex. Geological and evolution. Journal of Fish Biology 80: 918–951. Magazine 73: 488–503. Maisey JG, Anderson ME. 2001. A primitive chondrich- Moy-Thomas JA. 1939. Palaeozoic fishes. London: Methuen thyan braincase from the Early Devonian of South Africa. & Co. Journal of Vertebrate Paleontology 21: 702–713. Moy-Thomas JA, Miles RS. 1971. Palaeozoic fishes. 2nd ed. Maisey JG, Miller R, Turner S. 2009. The braincase of London: Chapman & Hall. the chondrichthyan Doliodus from the Lower Devonian Nelson GJ. 1968. Gill-arch structure in Acanthodes. In: Campbellton Formation of New Brunswick, Canada. Acta Ørvig T, ed. Nobel symposium 4, current problems of lower Zoologica 90 (Suppl. 1): 109–122. vertebrate phylogeny. Stockholm: Almqvist and Wiskell, Maisey JG, Naylor GP, Ward DJ. 2004. Mesozoic 128–143. elasmobranchs, neoselachian phylogeny and the rise of Nelson GJ. 1969. Gill arches and the phylogeny of fishes, modern elasmobranch diversity. In: Arratia G, Tintori A, with notes on the classification of vertebrates. Bulletin of eds. Mesozoic fishes 3 – systematics, paleoenvironments and the American Museum of Natural History 141: 475–552. biodiversity. Munich: Verlag Dr. Friedrich Pfeil, 17–56. Newberry JS. 1875. Description of fossil fishes. Report of the Marinelli W, Strenger A. 1954. Vergleichende Anatomie der Geological Survey of Ohio 2: 1–64. Wirbeltiere. 1. Lampetra fluviatilis. Vienna: Franz Deuticke. Nielsen E. 1932. Permo-Carboniferous fishes from East Marinelli W, Strenger A. 1956. Vergleichende Anatomie der Greenland. Meddelelser om Grønland 86: 1–63. Wirbeltiere. 2. Myxine glutinosa. Vienna: Franz Deuticke. Nielsen E. 1949. Studies on Triassic fishes from East Mark-Kurik E, Botella H. 2009. On the tail of Errivaspis Greenland. II. Australosomus and Birgeria. Meddelelser om and the condition of the caudal fin in heterostracans. Acta Grønland 146: 1–309. Zoologica 90 (Suppl. 1): 44–51. Obruchev DV. 1964. Class Placodermi. In: Orlov IA, ed. Märss T. 1979. [Lateral line sensory system of the Ludlovian Fundamentals of paleontology, 11. Agnatha, Pisces. Moscow: thelodont Phlebolepis elegans Pander.]. Eesti NSV Teaduste Nauka, 118–171. Akadeemia Tomimetised, Geoloogia 28: 108–111 (in Russian Ørvig T. 1957. Notes on some Paleozoic lower vertebrates with English summary). from Spitsbergen and North America. Norsk Geologisk Märss T. 1986. Squamation of the thelodont agnathan Tidsskrift 37: 285–353. Phlebolepis. Journal of Vertebrate Paleontology 6: 1–11. Ørvig T. 1960. New finds of acanthodians, arthrodires, Märss T, Turner S, Karatajute-Talimaa V. 2007. ‘Agnatha’ crossopterygians, ganoids and dipnoans in the upper II – Thelodonti. In: Schultze H-P, ed. Handbook of Middle Devonian calcareous flags (Oberer Plattenkalk) paleoichthyology, vol. 1B. Munich: Verlag Dr Friedrich Pfeil, of the Bergisch Gladbach – Paffrath Trough. (Part 1). 1–143. Paläontologische Zeitschrift 34: 295–335. Märss T, Wilson MVH, Thorsteinsson R. 2002. New Ørvig T. 1962. Y a-t-il une relation directe entre les thelodont (Agnatha) and possible chondrichthyan (Gnatho- arthrodires ptyctodontides et les holocephales? Colloques stoma) taxa established in the Silurian and Lower Devonian internationaux du Centre national de la Recherche of the Canadian Arctic Archipelago. Proceedings of the Esto- scientifique 104: 49–61. nian Academy of Sciences, Geology 51: 88–120. Ørvig T. 1967. Phylogeny of tooth tissues: evolution of some Miles RS. 1970. Remarks on the vertebral column and caudal calcified tissues in early vertebrates. In: Miles AEW, fin of acanthodian fishes. Lethaia 3: 343–362. ed. Structural and chemical organisation of teeth. London: Miles RS. 1973a. Articulated acanthodian fishes from the Old Academic Press, 45–110. Red Sandstone of England, with a review of the structure Ørvig T. 1972. The latero-sensory component of the dermal and evolution of the acanthodian shoulder-girdle. Bulletin of skeleton in lower vertebrates and its phyletic significance. the British Museum (Natural History): Geology 24: 113–213. Zoologica Scripta 1: 139–155. Miles RS. 1973b. Relationships of acanthodians. In: Green- Ørvig T. 1975. Description, with special reference to the wood PH, Miles RS, Patterson C, eds. Interrelationships of dermal skeleton, of a new radotinid arthrodire from fishes. London: Academic Press, 63–103. the Gedinnian of Arctic Canada. Colloques internationaux Miles RS. 1977. Dipnoan (lungfish) skulls and the relation- du Centre national de la Recherche scientifique 218: 43– ships of the group: a study based on new species from the 71. Devonian of Australia. Zoological Journal of the Linnean Pander CH. 1857. Ueber die Placodermen des Devonischen Society 61: 1–328. Systems. St. Petersburg. Miles RS, Young GC. 1977. Placoderm interrelationships Patten W. 1904. New facts concerning Bothriolepis. Biologi- reconsidered in the light of new ptyctodontids from Gogo cal Bulletin 7: 113–124. Western Australia. In: Andrews SM, Miles RS, Walker AD, Patterson C. 1965. The phylogeny of the chimaeroids. Philo- eds. Problems in vertebrate evolution. London: Academic sophical Transactions of the Royal Society of London B 249: Press, 123–198. 101–219. Miller RF, Cloutier R, Turner S. 2003. The oldest articu- Patterson C. 1982a. Morphological characters and homology. lated chondrichthyan from the Early Devonian period. In: Joysey KA, Friday AE, eds. Problems of phylogenetic Nature 425: 501–504. reconstruction. London: Academic Press, 21–74.

Patterson C. 1982b. Morphology and interrelationships of and jaws in the earliest jawed vertebrates. Nature 491: primitive actinopterygian fishes. American Zoologist 22: 748–751. 241–259. Rücklin M, Giles S, Janvier P, Donoghue PCJ. 2011. Pearson DM, Westoll TS. 1979. The Devonian actinop- Teeth before jaws? Comparative analysis of the structure terygian Cheirolepis Agassiz. Transactions of the Royal and development of the external and internal scales in the Society of Edinburgh 70: 337–399. extinct jawless vertebrate Loganellia scotica. Evolution & Peignoux-Devill J, Lallier F, Vidal B. 1982. Evidence for Development 13: 523–532. the presence of osseous tissue in dogfish vertebrae. Cell and Sahney S, Wilson MVH. 2001. Extrinsic labyrinth infillings Tissue Research 222: 605–614. imply open endolymphatic ducts in Lower Devonian Poplin C. 1974. Étude de quelques Paléoniscidés pennsyl- osteostracans, acanthodians, and putative chondrichthyans. vaniens du Kansas. Cahiers de Paléontologie (Section Journal of Vertebrate Paleontology 21: 660–669. Vertébrés). Paris: Éditions du CNRS. Sansom IJ, Davies NS, Coates MI, Nicoll RS, Ritchie A. Pradel A, Sansom IJ, Gagnier P-Y, Cespedes R, Janvier 2012. Chondrichthyan-like scales from the Middle Ordovi- P. 2007. The tail of the Ordovician fish Sacabambaspis. cian of Australia. Palaeontology 55: 243–247. Biology Letters 3: 73–76. Sansom IJ, Donoghue PCJ, Albanesi G. 2005. Histology Pradel A, Tafforeau P, Maisey JG, Janvier P. 2011. A new and affinity of the earliest armoured vertebrate. Biology Palaeozoic Symmoriiformes (Chondrichthyes) from the late Letters 1: 446–449. Carboniferous of Kansas (USA), and cladistic analysis of Sansom IJ, Haines PW, Andreev P, Nicoll RS. 2013. A early chondrichthyans. PLoS ONE 6: e24938. new pteraspidomorph from the Nibil Formation (Katian, Purnell MA. 1994. Skeletal ontogeny and feeding mecha- Late Ordovician) of the Canning Basin, Western Australia. nisms in conodonts. Lethaia 27: 129–138. Journal of Vertebrate Paleontology 33: 764–769. Purnell MA, Donoghue PCJ. 1997. Architecture and func- Sansom IJ, Smith MP, Smith MM, Turner P. 1997. tional morphology of the skeletal apparatus of ozarkodinid Astraspis the anatomy and histology of an Ordovician fish. conodonts. Philosophical Transactions of the Royal Society, Palaeontology 40: 625–643. Series B 352: 1545–1564. Sansom RS. 2009. Phylogeny, classification and character Regan CT. 1904. The phylogeny of the Teleostomi. Annals polarity of the Osteostraci (Vertebrata). Journal of System- and Magazine of Natural History 4: 329–349. atic Palaeontology 7: 95–115. Reif W-E. 1982. Evolution of dermal skeleton and dentition in Sansom RS, Gabbott SE, Purnell MA. 2013. Unusual anal vertebrates: the odontode regulation theory. Evolutionary fin in a Devonian jawless vertebrate reveals complex origins Biology 15: 287–368. of paired appendages. Biology Letters 9: 20130002. http:// Reis O. 1895. Illustrationen zur des Skeletts von Acanthodes dx.doi.org/10.1098/rsbl.2013.0002 bronni Ag. Abhandlungen der Senckenbergischen Naturfor- Sansom RS, Rodygin SA, Donoghue PCJ. 2008. schenden Gesellschaft 19: 49–64. The anatomy, affinity and phylogenetic significance of Reis O. 1896. Ueber Acanthodes bronni Ag. Morphologische Ilemoraspis kirkinskayae (Osteostraci) from the Devonian of Arbeiten, herausgegeben von Dr. G. Schwalbe 6: 148–220. Siberia. Journal of Vertebrate Paleontology 28: 613–625. Ritchie A. 1967. Ateleaspis tessellata Traquair, a non- Schaeffer B. 1968. The origin and basic radiation of cornuate cephalaspid from the Upper Silurian of Scotland. the Osteichthyes. In: Ørvig T, ed. Nobel symposium 4, Zoological Journal of the Linnean Society 47: 69–81. current problems of lower vertebrate phylogeny. Stockholm: Roemer F. 1857. Über Fisch- und Pflanzen- führende Almqvist and Wiskell, 207–222. Mehrgelschiefer des Rothliegenden bei Klein-Neundorf Schaeffer B. 1969. Adaptive radiation of the fishes and the unweit Löwenberg, und im Besonderen über Acanthodes fish-amphibian transition. Annals of the New York Academy gracilis den am häufigsten in denselben vorkommenen of Sciences 167: 5–17. Fisch. Zeitschrift Deutsche Geologische Gesellschaft 9: Schaeffer B. 1975. Comments on the origin and basic 51–84. radiation of the gnathostome fishes, with particular refer- Romer AS. 1933. Vertebrate paleontology. Chicago, IL: Uni- ence to the feeding mechanism. Colloques internationaux versity of Chicago Press. du Centre national de la Recherche scientifique 218: 101– Romer AS. 1945. Vertebrate paleontology. 2nd edn. Chicago, 109. IL: University of Chicago Press. Schaeffer B. 1981. The xenacanth neurocranium, with com- Romer AS. 1966. Vertebrate paleontology. 3rd edn. Chicago, ments on elasmobranch monophyly. Bulletin of the Ameri- IL: University of Chicago Press. can Museum of Natural History 169: 1–66. Romer AS. 1968. Notes and comments on vertebrate Schultze HP, Zidek J. 1982. Ein primitiver Acanthodier paleontology. Chicago, IL: University of Chicago Press. (Pisces) aus dem Unterdevon Lettlands. Paläontologische Rosen DE, Forey PL, Gardiner BG, Patterson C. 1981. Zeitschrift 56: 95–106. Lungfishes, tetrapods, paleontology, and plesiomorphy. Bul- Schultze H-P, Cumbaa SL. 2001. Dialpina and the charac- letin of the American Museum of Natural History 167: ters of basal actinopterygians. In: Ahlberg PE, ed. Major 159–276. events in early vertebrate evolution: palaeontology, phylog- Rücklin M, Donoghue PCJ, Johanson Z, Trinajstic K, eny, genetics and development. London: Taylor & Francis, Marone F, Stampanoni M. 2012. Development of teeth 315–322.

Shu D, Conway Morris S, Han J, Zhang Z-F, Yasui K, vertebrates and related problems of evolutionary biology. Janvier P, Chen L, Zhang XL, Liu J-N, Li Y, Liu H-Q. Beijing: Science Press, 87–119. 2003. Head and backbone of the Early Cambrian vertebrate Turner S. 2004. Early vertebrates: analysis from microfossil Haikouichthys. Nature 421: 526–529. analysis. In: Arratia G, Wilson MVH, Cloutier R, eds. Recent Sire J, Donoghue PCJ, Vickaryous MK. 2009. Origin and advances in the origin and early radiation of vertebrates. evolution of the integumentary skeleton in non-tetrapod Munich: Dr Friedrich Pfeil, 65–94. vertebrates. Journal of Anatomy 214: 409–440. Valiukevicius J. 1992. First articulated Poracanthodes from Smith MM, Coates MI. 2000. Evolutionary origins of teeth the Lower Devonian of Severnaya Zemlya. In: Mark-Kurik and jaws: developmental models and phylogenetic patterns. E, ed. Fossil Fishes as Living Animals. Tallinn: Institute of In: Teaford MF, Ferguson MWJ, Smith MM, eds. Develop- Geology, 193–213. ment, function and evolution of teeth. Cambridge: Cam- Van der Brugghen W, Janvier P. 1993. Denticles in bridge University Press, 133–151. thelodonts. Nature 364: 107. Smith MM, Coates MI. 2001. The evolution of verte- Vladykov VD. 1973. A female sea lamprey (Petromyzon brate dentitions: phylogenetic pattern and developmental marinus) with a true anal fin, and the question of the models. In: Ahlberg PE, ed. Major events in early vertebrate presence of an anal fin in Petromyzonidae. Canadian evolution: palaeontology, phylogeny, genetics and develop- Journal of Zoology 51: 221–224. ment. London: Taylor & Francis, 223–240. Vladykov VD, Kott E. 1980. Description and key to meta- Smith MM, Hall BK. 1990. Development and evolutionary morphosed specimens and ammocoetes of Petromyzontidae origins of vertebrate skeletogenic and odontogenic tissues. found in the Great Lakes region. Canadian Journal of Biological Reviews 65: 277–373. Fisheries and Aquatic Sciences 37: 1616–1625. Smith MM, Johanson Z. 2003. Separate evolutionary Wang N-Z, Donoghue PCJ, Smith MM, Sansom IJ. 2005. origins of teeth from evidence in fossil jawed vertebrates. Histology of the galeaspid dermoskeleton and endoskeleton, Science 299: 1235–1236. and the origin and early evolution of the vertebrate cranial Stahl BJ. 1999. Chondrichthyes III. Holocephali. In: Schultze endoskeleton. Journal of Vertebrate Paleontology 25: 745– HP, ed. Handbook of paleoichthyology, volume 4. Munich: 756. Verlag Dr Friedrich Pfeil, 1–164. Watrous LE, Wheeler QD. 1981. The out-group comparison Stensiö EA. 1925. On the head of the macropetalichthyids method of character analysis. Systematic Zoology 30: with certain remarks on the head of other arthrodires. 1–11. Publications of the Field Museum of Natural History, Geo- Watson DMS. 1937. The acanthodian fishes. Philosophical logical Series 4: 87–197. Transactions of the Royal Society of London B 228: 49–146. Stensiö EA. 1931. Upper Devonian vertebrates from East Wiens JJ. 2003a. Incomplete taxa, incomplete characters, Greenland collected by the Danish Greenland expedition in and phylogenetic accuracy: is there a missing data problem? 1929 and 1930. Meddeleser om Grønland 96: 1–212. Journal of Vertebrate Paleontology 23: 297–310. Stensiö EA. 1963. Anatomical studies on the arthrodiran Wiens JJ. 2003b. Missing data, incomplete taxa, and head. Pt 1. Preface, geological and geographical distribution, phylogenetic accuracy. Systematic Biology 52: 528–538. and organisation of the arthrodires, the anatomy of the head Wiley EO. 1979. Ventral gill arch muscles and the interrela- in the Dolichothoraci, Coccosteomorphi and Pachyosteo- tionships of gnathostomes, with a new classification of the morphi. Kungliga Svenska Vetenskapsakademiens Handlin- Vertebrata. Zoological Journal of the Linnean Society 67: gar 9: 1–419. 149–179. Stensiö EA. 1969. Elasmobranchiomorphi Placodermata Wilga C. 2005. Morphology and evolution of the jaw suspension Arthrodires. In: Piveteau J, ed. Traité de Paléontologie 4. in lamniform sharks. Journal of Morphology 265: 112–119. Paris: Masson, 71–692. Williams ME. 1998. A new specimen of Tamiobatis vetusus Takezaki N, Figueroa F, Zaleska-Rutczynska Z, Klein J. (Chondrichthyes, Ctenacanthoidea) from the Late Devonian 2003. Molecular phylogeny of early vertebrates: monophyly Cleveland Shale of Ohio. Journal of Vertebrate Paleontology of the agnathans as revealed by sequences of 35 genes. 18: 251–260. Molecular Biology and Evolution 20: 287–292. Williams ME. 2001. Tooth retention in cladodont sharks: Traquair RH. 1875. On the structure and systematic position with a comparison between primitive grasping and swal- of the genus Cheirolepis. Annals and Magazine of Natural lowing, and modern cutting and gouging feeding mecha- History 15: 237–249. nisms. Journal of Vertebrate Paleontology 21: 214–226. Traquair RH. 1888. Notes on the nomenclature of the fishes Wilson MVH, Caldwell MW. 1998. The Furcacaudiformes: a of the Old Red Sandstone of Great Britain. Geological new order of jawless vertebrates with thelodont scales, Magazine 5: 507–517. based on articulated Silurian and Devonian fossils from Trinajstic K, Long JA, Johanson Z, Young G, Senden T. northern Canada. Journal of Vertebrate Paleontology 18: 2012. New morphological information on the ptyctodontid 10–29. fishes (Placodermi, Ptyctodontia) from Western Australia. Wilson MVH, Hanke GF, Märss T. 2007. Paired fins of Journal of Vertebrate Paleontology 32: 757–780. jawless vertebrates and their homologies across the Turner S. 1991. Monophyly and interrelationships of the ‘Agnathan’-Gnathostome transition. In: Anderson JS, Thelodonti. In: Chang M-M, Liu Y-H, Zhang G-R, eds. Early Sues H-D, eds. Major transitions in vertebrate evolution.

Bloomington and Indianapolis: Indiana University Press, Zalc B, Goujet D, Colman D. 2008. The origin of the 122–149. myelination program in vertebrates. Current Biology 18: Woodward AS. 1891. Catalogue of the fossil fishes in the R511–R512. British Museum (Natural History). Part II. London: Trus- Zangerl R. 1981. Chondrichthyes I. Paleozoic Elasmobranchii. tees of the British Museum (Natural History). In: Schultze H-P, ed. Handbook of paleoichthyology, volume Woodward AS. 1932. Text-book of palaeontology, volume 2. 3A. Stuttgart: Gustav Fischer Verlag, 1–115. London: Macmillan. Zhang G-R, Wang J-Q, Wang N-Z. 2001. The structure of Yalden DW. 1985. Feeding mechanisms as evidence for pectoral fin and tail of Yunnanolepidoidei, with a discussion cyclostome monophyly. Zoological Journal of the Linnean of the pectoral fin of chuchinolepids. Vertebrata PalAsiatica Society 84: 291–300. 39: 1–13. Young GC. 1978. A new Early Devonian petalichthyid fish Zhu M, Janvier P. 1998. The histological structure of the from the Taemas/Wee Jasper region of New South Wales. endoskeleton in galeaspids (Galeaspida, Vertebrata). Alcheringa 2: 103–116. Journal of Vertebrate Paleontology 18: 650–654. Young GC. 1979. New information on the structure and Zhu M, Yu X. 2002. A primitive fish close to the common relationships of Buchanosteus (Placodermi: Euarthrodira) ancestor of tetrapods and lungfish. Nature 418: 767–770. from the Early Devonian of New South Wales. Zoological Zhu M, Yu X, Ahlberg PE. 2001. A primitive sarcopterygian Journal of the Linnean Society 66: 309–352. fish with an eyestalk. Nature 410: 81–84. Young GC. 1980. A new Early Devonian placoderm from New Zhu M, Yu X, Ahlberg PE, Choo B, Lu J, Qiao T, Qu Q, South Wales, Australia, with a discussion of placoderm Zhao W, Jia L, Blom H, Zhu Y. 2013. A Silurian phylogeny. Palaeontographica 167A: 10–76. placoderm with osteichthyan-like marginal jaw bones. Young GC. 1986. The relationships of placoderm fishes. Zoo- Nature 502: 188–193. logical Journal of the Linnean Society 88: 1–57. Zhu M, Yu X, Choo B, Qu Q, Jia L, Zhao W, Qiao T, Lu J. Young GC. 1991. The first armoured agnathan vertebrates 2012a. Fossil fishes from China provide first evidence of from the Devonian of Australia. In: Chang M-M, Liu Y-H, dermal pelvic girdles in osteichthyans. PLoS ONE 7: Zhang G-R, eds. Early vertebrates and related problems of e35103. evolutionary biology. Beijing: Science Press, 67–85. Zhu M, Yu X, Choo B, Wang J, Jia L. 2012b. An antiarch Young GC. 2003. Did placoderm fish have teeth? Journal of placoderm shows that pelvic girdles arose at the root of Vertebrate Paleontology 23: 988–990. jawed vertebrates. Biology Letters 8: 453–456. Young GC. 2008. Number and arrangement of extraocular Zhu M, Yu X, Janvier P. 1999. A primitive fossil fish sheds muscles in primitive gnathostomes: evidence from extinct light on the origin of bony fishes. Nature 397: 607–610. placoderm fishes. Biology Letters 4: 110–114. Zhu M, Yu X, Wang W, Zhao W, Jia L. 2006. A primitive fish Young GC. 2010. Placoderms (armored fish): dominant ver- provides key characters bearing on deep osteichthyan phy- tebrates of the Devonian Period. Annual Review of Earth logeny. Nature 441: 77–80. and Planetary Sciences 38: 523–550. Zhu M, Zhao W, Jia J, Lu J, Qiao T, Qu Q. 2009. The Yu X. 1998. A new porolepiform-like fish, Psarolepis romeri, oldest articulated osteichthyan reveals mosaic gnathostome gen. et sp. nov. (Sarcopterygii, Osteichthyes) from the characters. Nature 458: 469–474. Lower Devonian of Yunnan, China. Journal of Vertebrate Zittel KA. 1893. Grundzüge der Palæontologie (Palæozoologie). Paleontology 18: 261–274. Munich: Druck und Verlag von R. Oldenbourg.